-
The Language of the Buddhist Sanskrit Texts
John Brough
Bulletin of the School of Oriental and African Studies,
University of London, Vol. 16, No. 2.(1954), pp. 351-375.
Stable
URL:http://links.jstor.org/sici?sici=0041-977X%281954%2916%3A2%3C351%3ATLOTBS%3E2.0.CO%3B2-5
Bulletin of the School of Oriental and African Studies,
University of London is currently published by School of Oriental
andAfrican Studies.
Your use of the JSTOR archive indicates your acceptance of
JSTOR's Terms and Conditions of Use, available
athttp://www.jstor.org/about/terms.html. JSTOR's Terms and
Conditions of Use provides, in part, that unless you have
obtainedprior permission, you may not download an entire issue of a
journal or multiple copies of articles, and you may use content
inthe JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this
work. Publisher contact information may be obtained
athttp://www.jstor.org/journals/soas.html.Each copy of any part of
a JSTOR transmission must contain the same copyright notice that
appears on the screen or printedpage of such transmission.
JSTOR is an independent not-for-profit organization dedicated to
and preserving a digital archive of scholarly journals. Formore
information regarding JSTOR, please contact [email protected].
http://www.jstor.orgTue May 1 20:43:50 2007
-
THE LANGUAGE OF THE BUDDHIST SANSKRIT TEXTS
By JOHNBROUGH
WHEN Buddhist works in Sanskrit were first introduced into
Europe, it was at once obvious that the language of some of them as
it appears in the manuscripts was, in comparison with Classical
Sanskrit, frequently ungram- matical, and on occasion barbarously
so. The immediate and natural reaction of scholars accustomed to
the regularity of Sanskrit was to stigmatize these shortcomings,
and to attempt to remove as many irregularities as possible by
forcibly emending the text. I t was, however, very soon recognized
that many of the seeming anomalies could not be abolished, and that
they must be accepted as genuine in their own context. This was
especially clear in the case of the verses of some of the older
texts, where the metre often guaranteed non-Piiqinean forms ; and
the language of these verses, variously called the Giithii-
dialect, mixed Sanskrit, or hybrid Sanskrit, was recognized as
something in its own right. The same courtesy was readily extended
to the prose of the Mahi- vastu, which in places could only have
been made to resemble Sanskrit by completely rewriting the text.
The prose of the other texts, being in many ways virtually
Sanskrit, took considerably longer to win the same recognition ;
but for many years now it has been generally admitted that here
also is a language which must be judged according to its own
standards, and not exclusively by the canons of classical Sanskrit
grammar.
I t is possible, however, for an editor to accept all this in
principle, and yet to be in serious doubt when trying to establish
a text ; for unless he has a grammatical norm against which to
measure his text, he is unable to apply the diagnostic test of
grammatical abnormality, which in classical Sanskrit or in Latin
would often provide the first hint that a passage is probably
corrupt. Hitherto, editors have had to make shift with the
classical grammar and dictionary, supplemented by their own memory
of other Buddhist texts. But the lack of a systematic study of
Buddhist Sanskrit has frequently resulted in over-correction by
editors, and a considerable number of the published texts really
require re-editing.
In these circumstances it is a matter of great satisfaction to
all who are con- cerned in this field that Professor Franklin
Edgerton has now published a gram- mar and dictionary of Buddhist
Sanskrit.1 This is a major work, the fruit of many years' careful
study, and it must remain for a long time to come a vade mecum for
future editors of Buddhist texts. Indeed, it is probably not an
exaggeration to say that it may well determine for a generation the
attitude of young editors towards their texts. For this reason I
should like to discuss a number of questions arising out of the
work in rather more detail than is customary in a review. But I
would ask the reader to remember that, although some of these
1 Buddhist Hybrid Sanskrit Grammar and Dictionary ; and Buddhist
Hybrid Sanskrit Reader. Yale University Press, 1953. See also
below, p. 421.
VOL. XVI. PART 2. 25
administratorHighlight
administratorHighlight
-
352 J. BROUGH
matters are important, the total of my criticism concerns only a
relatively small part of the work ; and I should not wish it to be
thought that I am in any way lacking in appreciation of the great
value of the work as a whole, or in gratitude for the enormous
labour bestowed by the author on his task. Rather, most of what I
have to say is in the nature of a few additional footnotes and
adjustments, together with a few suggestions to indicate a possible
direction for future work in this field.
Both Grammar and Dictionary are confined to reporting forms and
words or meanings which do not occur in Classical Sanskrit. I t
follows that a reader would get a most distorted picture of the
language from reading only the Grammar, which, being a systematic
collection of anomalies, is serviceable only to one who already
knows the texts. There is of course no ground for complaint in
this. I t is in the result primarily a grammar of the ' gathl '
language, and it cannot claim to provide a complete grammatical
picture of the Buddhist Sanskrit texts as a whole. Indeed, it is
doubtful whether an editor of a Buddhist Tantra, or a medieval
verse Avadlna-text, would get very much direct help from the
Grammar, although the Dictionary u~ould of course be valuable. The
material is arranged in the Grammar according to the categories of
classical Sanskrit, and while this procedure is satisfactory as
providing a ready means of reference, it occasionally induces
explanations which seem to me unjustified. For example, nouns
ending in -a frequently occur in verses as plurals and the direct
object of verbs ; and it would doubtless be sufficient to say of
these that they are simply uninflected or stem forms (a situation
which is admitted for the singular in 8.3 ff.),l and that their
plurality and their status as direct object arise from the context.
To say that they represent a metrical shortening of -a, which is
itself a nominative plural used in the sense of the accusative, is
to tie the grammar into quite unnecessary knots (8.94). In the same
way it seems unnecessarily complicated to say that in anyatra karma
sukrtad the form karma is an ablative of an a-stem for an la-stem,
i.e, for *karmi(t), with metrical shortening of -a (8.9). The
alternative explanation simply as a stem-form, admitted in 17.13,
seems much to be preferred ; and it may perhaps be suggested that
in instances like this the inflexion of sukrtad was felt to belong
to the phrase as a whole.
As already remarked, many of the published texts are badly
edited, and Edgerton has supplemented them wherever possible by
making full use of manuscript variants when these are reported by
editors ; and in the course of the work he has many valuable
corrections and emendations to suggest. While the work will
naturally be very useful to those who simply wish to read and
understand Buddhist Sanskrit, its chief value will undoubtedly lie
in the fact that it will assist future editors to produce better
editions ; and it is chiefly from the point of view of editors that
the following is written.
As we have noted, there has always been a tendency for editors
to lean too much towards correct classical Sanskrit. Edgerton, in
reaction against this,
1 References in figures with no other indication are to the
sections of the Grammar.
administratorHighlight
administratorHighlight
administratorHighlight
-
353 THE LANGUAGE O F THE BUDDHIST SANSKRIT TEXTS
goes to the other extreme, and in the preface to the Reader he
propounds a principle for editors : ' Any non-Sanskritic form
presented in the MSS. must, in general, be regarded as closer to
the original form of the text than a " correct " Sanskrit variant
'. The term ' non-Sanskritic '-which would cover all sorts of
copyists' blunders-is modified a few lines later into ' Middle
Indic or semi- Middle Indic '. Even this, however, seems to me to
go much too far. On some occasions, which we shall note below,
apparently middle-Indian forms can very easily result from scribal
error ; and in some contexts (in the semi-kivya style, for example)
any markedly non-Sanskrit form would be highly improbable. We shall
return to this point later. Edgerton adds that this principle is
not to be applied mechanically ; that the context, as well as
variant manuscript readings, will vary from case to case, and each
must be separately studied. My fear is that, from excessive
reaction against earlier editions, editors may not take this caveat
sufficiently to heart, and that we may have a crop of bad editions
comparable to the notorious edition of the Xvayarnbhupuriqa, where
the Brahman editor, considering that Buddhists could not be
expected to write good Sanskrit, seems to have put into his text
deliberately numerous copyists' errors from the manuscripts.
It seems to me that Edgerton throughout rather underestimates
the degree of accidental transmissional corruption which our texts
may have suffered, and many of his notes seem to imply that at
least the archetype of our manu- scripts must be correct, except in
those places where more Sanskritic forms have been intentionally
introduced. I t may be that an unreasoning confidence in the
accuracy of the scribe's hand and eye is traditional in these
studies ; for in 1916 we find Liiders writing l: ' For sragsitavin
the Nepalese MSS. read sagiritavan. The correct reading undoubtedly
is sragsitavi~a, but it is difficult to understand how this should
have been replaced by sagiritavin, unless we assume that the
original reading was a Prakrit form, such as, e.g. sagsitavi. This
has been correctly Sanskritized into sragsitava?~ in the fragment,
whereas in the Nepalese version it was wrongly rendered by
sagiritavin '. This is incredible as an argument ; and indeed it
does not require much experience of Nepalese manuscripts to realize
that sragsitavin could hardly fail to be read and transcribed by
some copyist or other as sagiritavin, and that a recon- struction
from the Prakrit need not enter into the picture here at all.
Before it could have any force, Luders' argument would require
assent to the pro- position that all textual corruption is
interpolation, a dogma of scribal infalli- bility which few editors
would care to hold.
Some of Edgerton's conclusions seem to depend upon a rather
similar faith in the scribes, or at least in the scribe of the
archetype. By way of illustration I shall deal here with a few
matters of orthography.
Since most of our texts depend either exclusively or chiefly
upon Nepalese manuscripts, it is desirable to consider the
idiosyncrasies of Nepalese scribes in
In ~Tfanuscript remains of Buddhist Literature found i n E.
Turkestan, ed. A. I?. R. Hoernle, p. 161.
administratorHighlight
administratorHighlight
administratorHighlight
administratorHighlight
administratorHighlight
-
354 J. BROUGH
general before attempting to assess the credibility of their
witness in any particular instance. The main points of spelling are
well known, and a few are mentioned almost as a matter of course in
the introductions to editions, but as they seem to have been
accorded rather less weight than is their due, it may be useful to
give a brief account of them here. The following frequently inter-
change : i and S ; u and G, r and ri (seldom ra) ; e, ya, and ye ;
o, va, and vo ; j a and ya ; j'a and gya ; la and ta ; ra and la ;
i a and sa ; 8a and kha ; k8a, cha, khya, and occasionally kha. In
many of these, in particular r/l, 9/s, it would seem that Newar
scribes considered the two forms to be merely graphic options, to
be used haphazard according to the fancy of the moment, in much the
same way as the copperplate forms of s and capital T alternate at
random in my own handwriting with forms based on the printed shapes
of these letters. In addition to these options, the use of the
superscript r is of interest. Since the following consonant is
regularly doubled, a bond seems to have been established between a
double consonant and a superscript r, and as a result any double
consonant may attract to itself a superscript r. The alternations
of spellings with and without the r then would seem to have led to
its occasional use over other conjuncts and even over single
consonants, and to its equally frequent omission where it is
historically required ; and it is hfficult to avoid the impression
that the sign was felt to be a mere ornament of the
handxmiting-perhaps playing a similar prestige role to that of the
b in doubt or the c in scissors when these spellings were first
introduced into our own orthography. Most of the following
examples, illustrating the results of some of these spelling habits
and of a few others less frequent, come from the Cambridge
manuscript of the A8tamS-vrata- mihatmya, but similar forms are
frequent in most Newari manuscripts. The list includes only
Sanskrit borrowings (though the same fluctuations in spelling
appear also in Newari words), and the standard Sanskrit spelling is
almost always equally permissible.
vake (vakya) ; arbhigyaka (abhi~eka) ; sarvage (sarvajpa) ;
yojpakvara (yogei- vara) ; jagya (yajpa) ; jvajpa (yogya) ; dyervva
(deva) ; rvyara (vela, for Class. vela) ; rmEra (mira) ; rbhikgu
(bhlkpu) ; nErma (nama) ; urttama, urttarma, uttarmma (uttama) ;
burddha (buddha) ; vasirtha (vasiptha) surga (sukha) ; nimirtti,
nimisti, nimistri (nimitti) ; dullabha (durllabha) ; vEttE (vir t t
l ) ; jalma, jarlma, jarnma, jarmma, jamma, jartma, jatma (janma) ;
yarma, janma (yama) ; nilmala, nilmara, nirlma [perhaps intended to
be read as nirmla] (nirmala) ; likhi (I@); ZEjakura (rljakula) ;
kyd& (krida) ; dharmEtramE, dhatmtitramti (dharmatml) ; mokga
(miircha) ; bhavikga (bhavieya) ; chichirika [monk's staff: cf.
Dict. svv. khakkhara, khakhar/a(ka), khikkhira, and add to
references under khankharaka, Gilgit MSS., vol. iii, 2, p. 142, and
p. x ; given by Hodgson, Essays on the Languages, Literature, and
Religion of Nepal and Tibet, p. 141, as khikshari ; in the Sanskrit
text of the PEpa-parimocana, 27, as kiikgirikE, kcikgTrZ].
Now, although the above examples come from Newar scribes writing
Newari texts, almost all of these vagaries can occur when the same
scribes write Sanskrit
administratorHighlight
-
355 THE LANGUAGE OF THE BUDDHIST SAXSKRIT TEXTS
manuscripts. In most of the latter, however, such spellings are
decidedly less frequent than in Newari. This is an important point,
since it shows that the scribes did not learn such habits from
Sanskrit manuscripts, and it is therefore not possible to argue
that some of these aberrations are due to the form in which the
Sanskrit tradition had been handed down. The scribes apparently
were quite aware that there was a norm of Sanskrit spelling and
attempted for the most part to follow it when copying a Sanskrit
text. But even the best of them are liable on occasion to introduce
Newari spellings, and most Nepalese Sanskrit manuscripts show a
fair number.
In the light of these considerations, it appears that a number
of manuscript spellings quoted in Edgerton's Grammar are not in any
way evidence for the forms of the original texts. Thus, for
example, he says (1.32), ' The BHS occurrences of 1 for r are
balanced by a substantial number of r for I,' and in 2.49 he gives
a list of both changes : ankula, kala (for kara), Kubela, vicilana,
panjala, abhin-ira, kira (for kila), raghu, i-itara, sakara, etc.
Since no doubts are expressed, and since these forms are allowed to
appear in the Dictionary, it would seem that he accepted them as
genuine. But since any initial or intervocalic r or 1 may be
written on any given occasion as 1 or r respectively (less
frequently in conjuncts, though even here it occurs from time to
time), it is clear that spellings such as the above can give us no
information at all. This is so even if at any given point all the
available manuscripts agree.
A typical example is provided by the bird-name
karavinka/kalavinka. Both forms are entered in the Dictionary, and
it is, of course, recognized that both denote the same bird, since
they occur in identical stereotyped contexts. (Incidentally, it
seems unlikely that the Indian cuckoo is meant, since kokila, the
common name for the latter, occurs alongside kalavinka in lists of
bird- names.) Under karavinka the note is added : ' In meaning
=Pali karavT, -vika ; in form a blend of this with kalavinka, which
in Skt. = sparrow '. But Pali also has kalavinka, and it seems most
probable that we have here two dialectally different Middle Indian
forms of the same word, kalavinka and karavzka (which may, of
course, have subsequently been differentiated in meaning). In the
Buddhist Sanskrit texts, however, the citations in the Dictionary
present only karavinka twice from the Lalitavistara, and kalavinka
thrice. I should, therefore, have no hesitation in attributing the
variation simply to the scribes, and restoring kalavinka everywhere
in Sanskrit. If, on the other hand, a form in -vTka were to turn up
in a Buddhist Sanskrit text, then an editor might incline towards
the spelling karavika, on the basis of the Pali evidence (whether
his manuscripts had -1- or -r-) : though even here a doubt would
still be possible, in view of the normal scribal indifference con-
cerning both the vowel-length and the employment or omission of the
anusv6ra.
With regard to the sibilants, a doubt is indeed expressed (2.56)
that ' corrup-tions in the tradition are very much to be suspected
in this case '. But again, since virtually any s may be written as
S, and vice versa, the list in 2.63 cannot tell us anything. Even
the sporadic appearance of -it- for -st-, of which two
administratorHighlight
administratorHighlight
administratorHighlight
-
356 J. BROUGH
examples are quoted in 2.61, is not convincing, since the
interchange also appears optionally in Newari spelling, for example
asti and agti (Skt. asthi), both of which occur within a few lines
of one another in the Newari version of the Pcipaparimocana. If
therefore we accept the reading agtayga (see Dict. s.v.) for
astayga in LV 390.8, it will be only because of the first of
Edgerton's arguments, namely the play upon words with actam in the
following line, though it may be felt that this is rather slender
support for an isolated anomaly. (I should like to suggest here
that the entry which follows in the Dictionary, AgtabhuginZ, the
gotra of the nakeatra Revati, Divy. 641.11, may perhaps be emended
to Artabhcigini. This emendation would imply the con- verse of the
type of misspelling noted above in Vasirtha for Basi,stha, and
would produce a well-attested gotra-name.)
In the same way, the frequent aberrations of the superscript r
make it most improbable that adhivattati (2.11) is really an
example of assimilation. The form is quoted from one manuscript in
a single passage (Mv. i.269.15), and because it coincides in
spelling with the Pali form-quite accidentally, in my view-Edgerton
(Dict. s.v.) suggests that it should be introduced into the text.
But it seems unlikely that we have here an original reading which
other manuscripts have independently Sanskritized ; whereas it is
quite natural to suppose that the force of the r had resulted in a
double consonant in an older copy, and that the ' optional'
superscript had afterwards been omitted. Precisely analogous to
this is a spelling of a personal name, Dharnmasiyha, which I have
seen on an 18th century Nepalese bronze figure, where Dharnma- is
not to be directly connected with the Pali assimilation of -rm-,
but is a com- paratively recent orthographic variant for -rrnm-. In
the same way, there is no reason to consider dullabha to result
from an early assimilation of -rl- in durlabha: rather, it is
merely an alternative spelling for durllabha. Such spellings may
perhaps reflect Nepalese pronunciations, but they are not in
themselves sufficient evidence for the original texts.
-
The converse of this situation appears in the spelling marjjay,
marjay, for Skt. majjan, in Mv. i.20.2. Senart retains in his text
the form with -r-, but Edgerton here is rightly doubtful, remarking
that if the form is to be kept, it is hyper-Skt. In view of the
known propensities of the scribes in this matter, the -r- can be
banished from the text without hesitation.
A doubt likewise attaches to some of the forms in 2.34 ( jfor y,
and y for j). A single occurrence each of jakrt for yakyt (LV
208.18, where in any case a v.1. ya- is reported), and Yarnbhaka
for Jambhaka (Mv. ii.112.6) is hardly enough to justify their
acceptance. The spelling an6rjay for anaryay (Nv. ii.79.3) is again
typically Newari, and can hardly be accepted as valid on the basis
of this single appearance.
Forms such as tydhd, tyvidhay, rnyyati, are quoted as examples
of ' hyper-Sanskrit ' formations, and the description may be
accepted, provided that it is understood that the culprit is again
in all probability the scribe. Similarly, a ' hyper-Sanskrit '
spelling such as trikcutto has very little claim on an editor,
administratorHighlight
-
357 THE LANGUAGE OF THE BUDDHIST SANSKRIT TEXTS
and should doubtless be interpreted as trikhutto (Skt. -kytvaJ.
For the scribal variants in this word, see Dict., s.v. krtva).
I t seems to me that the normal confusion between e, ya, and ye
makes it quite impossible to be certain whether the feminine
oblique-case ending was -6ye or -6ya, or whether both were used by
the authors of the texts. Edgerton (p. 63), taking the manuscript
readings at their face value, remarks that their distribution among
the several texts is peculiar ; -aye is almost restricted to the
Mah&vastu where it is commoner than -6ya ; while the latter is
common in the verses of other texts, though -6ye also occurs. But
it is doubtful whether this really carries us beyond the spelling
habits of the scribe of the archetype of our Alahtivastu
manuscripts, which, as Senart recognized, must be of relatively
recent date. The alternative in -6e is much rarer than the other
two, presumably because the scribes were aware that this sort of
hiatus should not occur in a Sanskritic text. On one occasion where
it does occur, in the word imtie, Edgerton remarks that one
manuscript has imtiya : but for a Nepalese scribe this is the same
reading. There would certainly be no justification for accepting
-6e into the text, in spite of the fact that it is the normal
Prakrit form ; and the choice between -aye and -6ya can only be
certain when the quantity of the final syllable is metrically
determined.
An interesting example where this alternation e:ya:ye may be
applied to the interpretation of the text can be seen in SP. 209.5
(9.65). Here the Central Asian version has paraTpar6ya tatha
anyamanyam ; but the Nepalese recension has paranjpar6 eva
tathtinyarnanyarn, where paran~par6 is interpreted as an
instrumental. This is clearly intended to mend the metre, and to
get rid of the
-
hiatus in tatha anya- ; but it would be strange if this were
done only at the expense of introducing a new hiatus. I would
suggest that e is here written for ye, and an editor of the
Nepalese recension should, therefore, read paray- pardy' eva.
Closely linked to these orthographical questions are a few where
the chief consideration is palzographical. The most striking of
these is the acceptance by Edgerton of the forms ygiti, ytiti,
ylla. Of y t i t i (Dict.) he says, ' But for the repeated
occurrence one might suspect a merely graphic corruption for Skt.
jhat-iti (var. jhag-iti) ' ; and under ylla, ' Senart . . . was
inclined to think the word a graphic error for jhalla, as was
Burnouf ; but he kept the MSS. reading, which seems too common to
emend.' Here I think we ought to be much bolder, and reject all
these forms as monstrosities. The sequences yt-, y1- are, if I am
not mistaken, strangers to the normal phonological patterns of
Sanskrit ; and while this may be no fundamental objection for a
quasi-onomatopoeia, it would still be an extraordinary coincidence
if the Sanskrit pair jhatiti, jhagiti were matched by a Buddhist
pair ytiti, ygiti, and still more surprising if jhalla, already
provided with a partner in the jingle jhalla-malla-, should have a
synonym (which also usurps its place in the compound with rnalla)
distinguished from it in most Nepalese writing only by a single
short stroke. The repeated occurrence of these forms, which has
been relied upon to justify their reality,
administratorHighlight
administratorHighlight
administratorHighlight
-
358 J. BROUGH
means simply that jha- has repeatedly been misread as r-, either
by scribes or by editors. (The spelling rigiti in the
Mah6vyutpatti, if genuine, would show that scribes were quite
capable of this misreading.)
Equally suspect, I feel, are the 3rd plural optative and aorist
forms in -itsu(b), -etsu(b), etc. These again are accepted as real
by Edgerton, and indeed it would almost seem that he gives them
preference whenever they appear in the manuscripts. Two alternative
explanations are suggested (32.97, 98) : either -ensu(b) has become
*-elztsu(b), and then, with ' denasalization ', -etsu($); or the
singular in -et has engendered a plural -et-su, on the analogy of
aorists in -i, -i-gu. The second of these explanations would mean
that this form is in origin entirely distinct from those in
-eysu(b), -ensu(b) ; but if this is so, we are none the less forced
to admit that the scribes have so completely entangled it with them
that it would be a hopeless task to recognize it now with
certainty. The other explanation implies a historical development
which is admittedly possible ; and indeed there is evidence that it
did take place in one Prakrit dia1ect.l But it would be hazardous
to connect the present situation directly with this. I t is true
that the explanation permits -etsu(b) to remain historically linked
with -ensu(b) ; but it does not explain the apparently haphazard
alterna- tion in the manuscripts. Further, if -ns- had really
developed historically here to -ts-, it might seem probable that
only the latter ought to have survived ; or, if forms of different
ages are to be assumed as appearing in the same text, the
intermediate form in -nts- or -rgts- might have been reasonably
expected to occur also ; and so far as I am aware, it never does.
The scatter of the forms in the manuscripts shows beyond any doubt
that -ysu, -nsu, and -tsu are three ways of writing what to the
scribe was the same form ; and there is no reason to doubt that
they represent a single form for the authors of the texts also. If
this be accepted, then it seems that an editor must make a choice
between the -t- and the nasal. The choice between -rg- and -n- is a
different, and less import- ant question. But the fact that -rg-
does occur, coupled with the evidence from other Middle I n l a n
sources, decides absolutely in favour of the nasal. Thus, we shall
also accept abhdrgsu rather than abhdtsu, comparing (as Edgerton
does) the ASokan ahuysu ; and we shall emend tisitsu to tisirgsu
(or tisirgsu), comparing the Pali tisirgsu.
Edgerton, however, urges strongly that we must admit that the
author of the Mahtivastu actually used the form in -etsu(b). ' As
to -etsu(b) ', he writes (32.96). ' I cannot believe that the
hundreds of occurrences in Mv are all manu- script corruptions, as
Senart assumes. Why should copyists introduce second- arily such a
monstrous-seeming form, in such a regular and constant way ? '
See. T. Burrow, The Language of the Kharogthi Documents from
Chinese Turkestan, p. 19 (e.g., maytsa for miLysa) ; also the
Prakrit Dhammapada, ed. H. W. Bailey, BSOAS., xi, mhich has
satiana, satbara, ahitba for saysanna, saysdra, ahiysb respectively
; and also bhametbu (previously read by Senart bhamerjsu), mhich is
strikingly like the forms under discussion here. The Pali version
(Dhp. 371) has bhuvassu, which editors have emended to bhamassu, on
the basis of the Prakrit passage. But the whole situation here
appears still to await a satisfactory explanation.
administratorHighlight
-
THE LAFGUAGE OF THE BUDDHIST SANSKRIT TEXTS 359
This is a most dangerous argument, and while the task of editors
would certainly be much easier if its implications could be
accepted, it attributes to the scribes a degree of literal
trustworthiness far beyond their deserts. I t is true that the
manuscripts of a given text provide our primary evidence, but none
the less they must always be read in conjunction with all the other
information at our disposal, and in particular the knowledge
derived from the rest of the literature, and the knowledge of
scribal habits derived from manuscripts of other texts. Edgerton's
argument, if rigidly applied, would in some texts force us, for
example, to print nisphala for .izigphala, since the spelling with
the dental -s- is, within my experience, almost universal in
Nepalese manuscripts, though it is not mentioned by Edgerton in the
grammar, and I do not remember having met it in an edition. The
same reasoning would compel the adoption in other texts of jatma
for janma, since some manuscripts know of no other writing of the
word. In practice, editors restore nigphala and janma as readily as
they differentiate b and v in their texts, although Nepalese (and
many other northern) manuscripts distinguish them not a t all.
And as for the form -nsub itself, it is possible to point to one
instance where the stenlma codicum shows beyond doubt that a whole
group of manuscripts have the spelling -tsub ' introduced
secondarily ', namely Suvarqabhasottama- siitra, p. 241.6. Here the
manuscripts ACF have abhistavitsub, while BDE read -stavinsub, and
G shows -stavimu, which is clearly a miscopying of -stavi(q)su.
This is one of the few texts where, thanks to the very careful
edition by Nobel, the stemma is crystal clear. Denoting lost
manuscripts by Greek letters, it is as follows :
Archetype
If the readings quoted are viewed against this stemma, it is
virtually certain that the archetypal reading was -nsu(b) or
-qsu(b)-at all events, with the nasal-and that the alteration to
-tsu$ was introduced secondarily some- where in the neighbourhood
of the point /3. Any other explanation would involve coincidences
so improbable that they need not be considered.
The explanation here of the ' regular and constant way ' in
which -tsu($) is written in these forms may perhaps be sought more
in graphic than in
administratorHighlight
-
360 J. BROUGH
linguistic considerations. In some forms of the Central Asian
scripts, for example, the appearance of n and t is very similar ;
and it may well be that a t some stage in the development of
Nepalese writing there was a genuine coalescence in graphic forms
of conjuncts such as -ts- and -ns-, -tm- and -nm-. If this is so,
then it may be that, in the older Nepalese manuscripts at least,
the shapes which from their appearance we transcribe as -ts-, -tm-,
were actually intended by the scribes as -ns-, -nm-. Alternatively,
the confusion may simply have started through straightforward
misreading of archaic exemplars. How-ever this may be, it is
certain that some later scribes considered that they had two
alternatives for their free choice. Where the second member of the
conjunct is m (graphically close to s), this alternation has
clearly been assisted by the normal sandhi of t before a following
m. In a number of places I have seen phrases like tan me, with a
tail added to the n in a second hand, thus producing tat me. I t
would seem that a reader of the manuscript, either for his own
reassurance or in teaching a pupil, has ' restored ' the basic
grammatical form. From instances of this sort, some scribes may
have even derived the feeling that the perverse writing looked more
learned than the other, and for this reason introduced it elsewhere
also. The following examples illustrating the confusion in both
directions are taken from a single manuscript (K) of the
Xubhatita-ratna-karaqda of Arya-i~ra: verse 64, jatmatya-loke (for
yan martya-) ; 71, irzman sukha- (for irimat) ; 80, dnmd (for atma)
; 123, bh5tCt samCivEsayate (for bhztan, acc. plu.) ; 129, mathat
satva- (for mathan, acc. plu.) ; 167, nEnmakymena (for
niitmakrameqa) ; 174, anmanam (for atmanam) ; 194, syin kuryat (for
sylt). The last example shows that the alternation of t and n is
not confined to conjuncts with -s and -m, though most frequently
found with these. Another manuscript of the same text has in verse
107 tan pratra- for tat pitra-. This work, I should add, is in good
classical Sanskrit, and no editor could in any case dream of
accepting into his text such forms as mathat or bhztlt as
accusative plurals.
In view of all this, and in spite of the frequency of -etsu(b)
in the manuscripts of the Mahivastu, I should without hesitation
follow Senart in rejecting these forms entirely. Likewise,
vihatsyase (31.24) should be emended to vihansyase (or
vihavpyase).
Nepalesg manuscripts are decidedly poor witnesses for anusvfira
and visarga. These are readily dropped or inserted in the wrong
places, even in texts which are indisputably in real Sanskrit ; and
it seems therefore that an accusative singular in -a for -av need
not necessarily be original, unless justified by metre. The same
doubt would be in order with regard to some of the apparent acc.
pl. masc. forms in -as, etc., since out of eleven examples cited
(8.93), seven are
If the writing is careful, they ought not to be confused. In the
manuscript of the Kalpami-maqiitika (ed. with selected facsimiles
by Luders, Kleinere Samkrit-Texte, ii), the distinction between tm
and nm is perfectly clear, and an editor could hardly be forgiven
for mistaking one for the other. This, however, does not mean that
a scribe, who did not necessarily understand what he was writing,
might not on occasion have slipped even in copying an exemplar as
clear as this.
administratorHighlight
administratorHighlight
-
THE LAVGUAGE OF THE BUDDHIST SAXSKRIT TEXTS 361
followed by c- or t-, and may equally well intend, for all the
manuscripts can tell us, -%pi c- and - 6 ~ s t-. The two dots, as
used for the visa~ga, are fre- quently employed in Newari
manuscripts simply as a mark of punctuation, and this usage has
occasionally found its way into Sanskrit manuscripts also. A
distinction is sometimes made by writing the visarga as two small
circles, or in the shape of a figure 8 ; but a confusion is
possible, and editors should be on their guard. As an example, I
may quote the apparent form Erabhyab, a-hich occurs thrice in the
Cambridge manuscript of the Pcipaparimocana for the absolutive
Crabhya. The other manuscripts, however, supply the normal form ;
and although Edgerton's principle would incline us to accept
Crabhyab (we should then doubtless explain it as a
hyper-Sanskritism), I have little doubt that it arose simply from
an earlier manuscript a-hich used the two dots with the force of a
comma.
Palzeographic considerations may also perhaps be called on to
help to explain the strange form bhitatka, physician, about a-hich
Edgerton is rightly doubtful (2.38). The form occurs thrice in the
edition of the SP ; but else- where it appears as bhitagka, which
is also the form reported from the Paris manuscript. Edgerton adds,
' One cannot help wondering where Kern and Nanjio got their reading
bhitatka, allegedly found in all their MSS.'. I suspect the answer
lies in the fact that in many hands g in the conjunct closely
resembles t and that the manuscripts did in fact read bhitagka. If
it be not too hazardous, I would tentatively suggest that this may
in its turn have arisen from a misunderstanding of a where the dot
denotedreading h~, not the nasal, but the doubling of the
consonant, so that the correct reading would be bhitakka, formally
identical with the Pali. This use of the dot to show a doubled
consonant is familiar in Prakrit in South Indian manuscripts, but i
t
-
seems to have occurred sporadically in the North also. A good
example appears in the DhvanyCloka, ad iii, 36, where the editions,
with the northern as well as the southern manuscripts, read targsa,
though tassa is clearly required.
A word should be said here about the frequent use in the Grammar
of the argument : ' so the majority of the MSS.'. This is a ghost
which refuses to be laid, in spite of the efforts of generations of
critics. Only if the manuscripts are related in descent in quite
specific ways is a simple majority good evidence for the archetypal
reading. An example to the contrary is provided by the
Suvarq,abhCsottama-sutra,where the six manuscripts ABCDEF are all
descended from an interpolated copy (see the stemma given above),
and when they are united in opposition to G, which is independent
and relatively free from inter- polation, they are as often as not
wrong. On the other hand, the agreement of G with any one of the
interpolated group, even against the united testimony of all the
five others, is very weighty indeed ; and in such a case the
majority is almost certainly wrong. An example of this (unimportant
in itself) may be seen on p. 52 of Nobel's edition, where F and G
read imu, and all the others ima. Edgerton, on quoting the rare
dual form imu from this passage (8.75), appears to cast doubt on it
by adding the note : ' but the majority of MSS.
-
362 J. BROUGH
ima '. None the less, the agreement of these two manuscripts so
much out- weighs the remainder that there is an extremely strong
presumption that the archetype read imu, which Nobel accordingly,
quite properly, accepts into his text. It would, of course, be
perfectly in order for an editor to go on from here, and argue
that, for such and such reasons, the archetypal reading itself was
corrupt ; and (as a purely hypothetical case) he might even find
cause to believe that the correct reading was in fact ima. But if
he did, the one argument he could ?tot use is that it occurs in
ABCDE, which could only show it by several lucky scribal
emendations. The twin ghost, the ' best manuscript ', appears
equally frequently ; but it would be otiose at this date to
reiterate Housman on this point, and I content myself with the
single observation that in any given place the ' best ' manuscript
may be wrong, and the critic must decide without reference to this
label, which is, after all, only attached to the manuscript by
other critics.
In the foregoing paragraphs I have tried to demonstrate a number
of points on wliich the uncorroborated evidence of Nepalese
manuscripts is inconclusive. I do not, of course, mean to suggest
that it is proved that all the examples quoted of non-Sanskritic
forms are necessarily wrong. But whenever these forms can be shown
to be capable of resulting from common Nepalese scribal practice, I
feel that it would be foolhardy of an editor to attribute them to
his text, unless they can be supported by evidence from sources
other than Nepalese. As Edgerton rightly stresses, his principle is
not to be applied mechanically, but the editor must use his
judgment in every individual case. This being so, however, the
principle itself may well seem to be superfluous. Now i t is clear
that for the bulk of the more Sanskritic part of the Buddhist
writings, the prime criterion against which to measure Nepalese
manuscripts must of necessity be classical Sanskrit itself. If in
view,of Edgerton's work this statement appears reactionary, I
should qualify it in the same manner as he : that the measure is
not to be applied mechanically. But we must always have in the
centre of our consideration the fact that the authors of a very
large part of the Buddhist Sanskrit texts extant really did intend
to write Sanskrit. An editor must not, of course, if he can help
it, attribute to his author better Sanskrit than the latter wrote.
But it is sometimes an equal danger to underestimate the author's
Sanskrit ability.
It must be freely admitted that in many cases there can, from
the nature of the evidence, be no absolute certainty. The great
value of Edgerton's work is that it now enables us to see at a
glance those non-Sanskrit features which occur sufficiently
frequently, over a sufficient range of texts, for us to believe
that these features were accepted as part of the language by the
Buddhist authors themselves. The exceptions are those outlined
above, where constantly recurring features are much more probably
to be laid at the door of the scribes. We should thus, for example,
reject jatma and -etsub, because tm and nm, ts and ns, are so
frequently confused in writing, even in word-junctures ; but on the
other hand, I feel we ought to accept the form dhEtvEvaropaqa
(for
administratorHighlight
administratorHighlight
-
THE LANGUAGE OF THE BUDDHIST SANSKRIT TEXTS 363
dhitv-avaropaqa), though Edgerton rejects it : since the
manuscripts in more than one text are unanimous for the long
vowel,' and-the important point- the interchange of a and 6 is
sufficiently infrequent to make it unlikely that the manuscripts of
different texts should have chanced upon the same mistake.
In the absence of this sort of constancy of occurrence, isolated
anomalies must always remain uncertain. This is not to say that
they are necessarily wrong-;rat heydpeva are not logically
impossible. But if a simple graphical explanation can derive a
given rarity from a normal form, an editor must give this fact due
weight. Thus, while only two examples are quoted (4.26) of the
elision of final -i, both from verses of the Lalitavistara, there
seems to be no very strong reason for doubting them. But on the
other hand a form such as y6durbhEmi, reported from only one place
(p. 224 and 24.11) is immediately suspect, in spite of the
antithesis with anturdhami, since it could so easily be a scribal
misreading for -bhomi or a scribal omission of the single sign a in
-bhav6mi. Of other isolated forms, bhihi, which is quoted only
twice from the Mahivastu, seeins in itself to be a not impossible
popular formation (though Edgerton's explanation of it in 30.10 as
an aorist injunctive is complicated and improbable : if real, it
could hardly have been felt as anything other than an imperative).
But since elsewhere in the same text, and also in the Lalitavistara
we find bhahi, and since the indicative bhii is known in Prakrit,
an editor might well consider the possibility that bhShi is merely
a misreading for blzahi by the scribe of the archetype. Still less
credible is uvacat (33.10) which is quoted only once, from
MahEvastu iii.337.13. Edgerton notes that a variant
-
uvEca is also recorded here; but he apparently accepts the weird
reading and adds, ' the aor. avocat no doubt helped to create the
form '. I should be most reluctant to accuse any ancient Indian
author of this, and the monstrum horrendum informe is without doubt
the product of mere scribal corruption. Senart prints it in his
text, saying in his note that he had decided not to suppress this
hybrid form, but adding that it was in all probability a mistake.
But there is no chance at all of its being correct. Direct speech
is very frequently indeed introduced in the Mahivastu by the words
e t d avocat, and only slightly less often by etad uvEca. I t seems
almost certain that the text here originally had etad avocat, and
that at some stage in the descent a scribe, copying mechanically,
misread the o and wrote etad avacat. A corrector then, overlooking
the final t , and thinking that uvEca was meant, wrote in a u under
the -d. he resulting form has then been copied mechanically by some
of our manuscripts, while others have made the obvious scribal
correction to uvEca.
By far the most valuable supplement to the Nepalese information
is provided
To Edgerton's references from the Kdragia-uytiha, Dict. s.v.
auaropaqa, can now be added DudviyBaty-av&na, x. Note, however,
that in the SubhiLtita-ratna-karaqiaaround which the Dud. is built,
the separate manuscripts have in the colophon here dhiLtvdropaga-,
and the text has only the phrases dhiLtur dropyate, dhiLtoj
sarmiropaga-, dhiLtum ciropya. We may then say that for Buddhist
kcivya in correct Sanskrit the form ciropaga only seems to be used,
and that the auadcina form with dua- results from a contamination
of the other with ava-.
administratorHighlight
-
364 J. BROUGH
by the manuscripts of Buddhist texts found in Central Asia.
Indeed, it was largely due to them that scholars first started
seriously to investigate Buddhist Sanskrit as such. I t was on the
basis of fragments of a Central Asian version of the
Saddharrna-puqdarika, differing in numerous details from the known
Nepalese recension, that Luders suggested that many of the Sanskrit
Buddhist texts started off their career in a much more Prakritic
form, and that they had subsequently been gradually Sanskritized by
generations of scribes, in differing ways in different scribal
traditions. Edgerton's principle of preferring the less Sanskrit
reading was enunciated by Luders with reference to such texts ;
though from this point of view it ought to apply only to the choice
between the readings of the two recensions, andnot simpliciter to
the variants in the Nepalese manuscripts among themselves. And on
many of the points of orthography which are apt to vex an editor,
the Central Asian manuscripts, where they present a correct
Sanskrit form in opposition to a Nepalese spelling, undoubtedly
justify the former ; and it would be a mistake, I feel, to hold
that in these matters also the Sanskrit form in general is a '
correction '. I t is a great misfortune for Buddhist philology that
so little of the vast
literature translated from Sanskrit into Tibetan and Chinese has
survived in the original language ; and although the considerable
fragments which have been recovered from Central Asia in the last
half-century do provide us already with a very good sample of
manuscripts much earlier than the Nepalese, each additional text
which comes to light may furnish collateral evidence of great value
for the editor who has to work from Nepalese sources. The volumes
so far published of the texts from Gilgit already show much
interesting material, though unfortunately they reached Edgerton
too late to enable him to utilize them to the full. I t is to be
hoped that in due course photographic facsimiles of these
manuscripts will be published, or, if this be too expensive, that
at least microfilm copies might be supplied to the principal
libraries, so that their evidence might be exploited to the best
advantage. Meanwhile Professor Ernst Waldschmidt continues the
invaluable work
begun by Liiders of editing the Buddhist Sanskrit fragments from
Turfan ; and I take the opportunity here of noting a few points
relevant to our present purpose from two of the most recent of
these publications, the Mah6parinirvaqa- sctra (MPS) and the
Mah&vad&na-siitra (MAS).l The most important general
observation is that these canonical texts seem to show distinctly
less Prakritic tendencies than, for example, the contemporary
fragments of the Saddharma- puqdarika. I t is true that Prakritic
forms do occur, for example, devanukarnpita- ppotalj (for
-purupzlj) &IPS, p. 13 ; vipakyisya &lAS, pp. 15, 23
(Waldschmidt wrongly emends to vipakyinas on p. 15) ; janetrs (for
janayitri), ibid., pp. 21,33. These however are virtually confined
to the verses (though janetr5 on p. 21, 25, line 3, may perhaps be
in prose) ; and metre occasionally indicates a Prakritic
pronunciation, e.g. MAS, pp. 29, 30, bhavati, several times scanned
as two syllables ; MPS, p. 13, bhavati, two syllables, kalpayati,
three syllables :
See also below, p. 421.
administratorHighlight
-
THE LANGUAGE OF THE BUDDHIST SAXSKRIT TEXTS 365
though in the same verses pcjayati and bhojayitva must each have
their full four syllables. This last circumstance might perhaps
indicate that verses of this sort are not the result of a simple
transposition of a Middle Indian original into Sanskrit spelling.
And in spite of the non-Sanskritic features the verses of these
texts are in general almost as Sanskritic as the prose. Although an
occasional metrical shortening may occur, e.g. MAS, p. 17, mEtE
mahivu7ya prabhikarasya (note also that pr- does not make position:
Waldschmidt wrongly emends to mahmayC), none the less, the most
striking of the features of the hybrid glthas of the Lalitacistara
are absent, and it seems hardly likely that we have here a
Sanskritization of an earlier form in hybrid Sanskrit. Now it is
clear that the existence of these texts in relatively correct
Sanskrit
already as early as the sixth century A.D. (in some cases even
earlier) carries considerable weight ; and if our Nepalese
manuscripts in opposition to them show in a given place a
non-Sanskritic form, then it seems that, other things being equal,
there is a prima facie case for considering the latter to be a
corruption. To take a single example, the occurrence of the
spelling Gydhrakcte in &IPS, p. 7, will justify the restoration
of this form in the majority of the texts, even although the
Nepalese manuscripts in most cases favour the (apparently)
semi-Prakritic form GyddhakQte. I do not of course mean to suggest
that the Central Asian manuscripts are in
themselves always better than the Nepalese. Indeed, they are
frequently careless in detail, and sometimes perverse in a manner
comparable to the Nepalese writing of -tsub for -nsub. For example,
we find a scribe whose script clearly distinguishes b and v writing
forms such as vuddha and vodhisatva,l while another writes sarvba,
p i i r ~ba ,~ Similarly we find and a third ~ a t b a . ~
occasional confusion of i and 2,though not so frequently as in
Nepalese manu- scripts, e.g. nyaiidat, sukhaiii (for -EL),
eva~gvidha.~ Visarga is frequently dropped, particularly, it would
seem, at the end of verses ; but as it also fre- quently appears in
the correct places, its omission is in all probability due for the
most part to scribal carelessness. In one instance at least,
drakiyata for -tab, 3rd person dual, the omission could not be
attributed to a Prakritic original. In spelling conventions, we
already find at this early date the typical -8rp
for in , and ns for ~ g s ; and also 9s' for 799, e.g. viyiati,
trivs'at, MAS, pp. 14, 15, together with the normal spelling with ~
g . Of interest also is yanv aharp, MAS, p. 22 (for yan nv ahaTg),
which is very common in Nepalese manuscripts and is adopted by some
editors, though Edgerton does not mention it. Although it is
doubtless as vulgar as the spelling ' alright ' in English, its
occurrence as early as this clearly gives it as much right to be
considered by editors as the comparable simplifications in satca,
etc., which are likewise common both here and in Nepalese sources.
On the other hand, the Central Asian manuscripts do not seem to
show the
typical Nepalese weakness in the confusion of r and I ; and if
in individual 1 Kleinere Sanskrit-Tezte ii, pp. 203, 207, 208.
ibid., pp. 29,30. 3 Hoernle, -Manuscript Remains, pp. 133 f. MAS,
pp. 13,27,39. MAS, p. 35.
-
366 J. BROUGH
words they should disagree with classical Sanskrit in these
letters it is probable that their evidence should be accepted. Thus
sakara, depending only on Nepalese sources, should be rejected
(Edgerton, Dict. s.v., gives only one reference ; but in Newari
manuscripts this spelling is almost as common as sakala) ; vichlana
remains doubtful, in spite of the Ardha-MCgadhi form viyalaqa,
since the spelling could equally well have arisen in Nepal ; but
samprad6layati can safely be accepted, since, in addition to the
Pali sampaddeti, we can cite in its favour samprad6lya from a
Central Asian manuscript, against the standard Sanskrit
pradtirayati. Similarly, the Middle Indian liikha, liiha (Skt.
riikta) has its initial justified by the occurrence of liiha- in
MPS, p. 93. A spelling of great interest appears in the word
ca~paka and the related town-
name caqpE, MPS, pp. 31, 33, 57. This is most striking and not
at all the sort of thing which one would expect to arise simply
from textual corruption ; and it is therefore surprising that
Waldschmidt emends both, without comment, to the standard Sanskrit
ca~npaka, campa. I t should be noted that not only do both
manuscripts which preserve the passage show caqpaka here, but that
the manuscript of the Kalpana-maqditika (early 4th century A.D.)
also has the same spelling.1 On the latter passage Liiders
commented that he believed that we might recognize this to be the
earlier form, which later by assimilation became campaka. There
seems to be no room for doubt here. Burrow has pointed out that the
Tamil form of the word is caqpakam, ceqpakam ; and although Gonda
has argued for an Austro-Asiatic origin, these Buddhist occurrences
of the Tamil-like spehng seem to establish with certainty that,
whatever the ultimate derivation, i t was a Dravidian language
which was the immedate source from which the Sanskrit word was
borrowed.
Thus the Central Asian manuscripts, fragmentary though many of
them are, provide a most valuable supplement to our knowledge of
the Buddhist Sanskrit texts, not only in the new material which
they have brought to light, but also by helping us to assess more
justly the worth of the Nepalese tradition ; and this assistance is
no less welcome to an editor where they support the latter, as in
fact they do to a considerable degree. I t is true that in a text
like the Saddhrma-puq4ar;iX.a the Nepalese recension is convicted
of having to some extent corrected non-Sanskrit forms; and the same
tale is told of the SuvarqabhGsottarna-siitraby the fortunate
chance of the survival of the old palm-leaf manuscript G. But this
is only part of the picture. It is equally important that the
canonical fragments show us a t this early date a language which is
virtually the same as that presented by the Nepalese manuscripts
of, for example, the PrajnEpGramitEs, or the great Avadana
collections. I t would therefore be over-hasty, I feel, to conclude
that the whole range of the old Buddhist Sanskrit texts has been
equally subjected to a continuing process of Sanskritization. We
have, indeed, direct evidence to the contrary in the
1 Ed. H. Liiders, Kbinere Sanskrit-Texte, ii, p. 139. ibid., p.
39. 3 Transactions of the Philological Society, 1946,p. 17.
Bijdragen tot de Taal- , Land- , en Volkenkunde, 105,1, pp. 137
ff.
administratorHighlight
administratorHighlight
-
THE LANGUAGE OF THE BUDDHIST SANSKRIT TEXTS 367
preservation of the MahEvastu itself; since any scribal
corrector who knew enough Sanskrit to feel that the language of
the- Mahavastti required improve- ment would surely have been able
to produce a better ' corrected ' version than that which has in
fact come down to us. Also, as is well known, the distance of the
language of this text from Sanskrit varies considerably from one
place to another, and there is no good reason why a scribal editor
should not have produced a more uniform result. The simplest
explanation would certainly seem to be that these differences
represent the styles of different authors, possibly of different
ages, but that in essentials they have been transmitted in the form
in which they were left by the original compilers who built up the
illahavastu largely out of inherited materials.
Similarly, in the case of the Sanskrit canon, it is obvious from
comparing the Pali version that it is very largely constructed out
of older material in some Prakrit dialect ; but there seems to be
no reason for assuming that it is anything other than a quite
definite translation into Sanskrit, done at a specific period, when
the Sarvastiviidins decided to adopt Sanskrit as their official
language.
In direct opposition to this view Edgerton writes in the preface
to his Reader : 'All BHS texts, even the Mahavastu, have been
subjected to a good deal of Sanskritization, some of it very likely
going back to the original composi- tion of the work, but much of
it, in the case of most if not all BHS works, introduced by
copyists and redactors in the course of the tradition '. I t is
impossible to deny that this is true of some texts ; but put in
these terms it seems to me very much to overstate the case, and for
much of the extant Buddhist Sanskrit, if not the major part, it
would be nearer the truth to reverse
. -
the statement, and say that some degree of additional
Sanskritization has doubtless been carried out by scribes, but that
a very great deal of the Sanskritic appearance is in all
probability due to the original authors or compilers. I cannot
believe that the texts as we now have them in the manuscripts would
show such clearly defined distinctions of style if scribes or late
redactors had tampered with them to the extent which Edgerton seems
to envisage.
o n the basis of the degree of approximation to Sanskrit ~ d ~ e
r t b n classifies his material into three groups (Grammar, p.
xxv), in which (1)both prose and verse are hybridized-principally
the MahEvastu ; ( 2 )the verses are hybridized, but the prose has
relatively few signs of Middle Indian phonology or morphology, e.g.
Saddharrna-pu~larika, Lalitavistara, etc. ; (3) both prose and
verse are substantially Sanskritized, e.g. DivyEvadEna, etc. Of the
third group he says, ' Non-Sanskritic forms are not common ; the
vocabulary is the clearest evidence that they belong to the BHS
tradition '. This seems quite satisfactory ; but throughout most of
the work the term Buddhist Hybrid Sanskrit seems to be directly
applied to the language not only of group 1and the verse of group 2
, but equally to the prose of group 2 and to group 3. This is a
very different matter from saying that the latter are ' in the
tradition ' of BHS. I t is true that they share many features with
the hybrid texts, particularly in vocabulary and
VOL. XVI. PART 2. 26
administratorHighlight
-
368 J. BROUGH
syntax. But in spite of this, the language in the typical prose
style of the canonical works and the Avadinas is much further
removed from the more Prakritic portions of the Mahiivastu (see,
for example, the passage quoted below) than it is from Classical
Sanskrit. And if it is misleading to call the Avadtina-style simply
Sanskrit, it seems to me all the more misleading to group it with
the Githa-language as Hybrid Sanskrit. No one would deny that the
Avadiina-style has its own idiosyncrasies which mark it off from
anything Brahmanical. But to call it ' hybrid ' for this reason
seems as little justified as it would be to call a medieval Hindu
commentary ' Brahmanical Hybrid Sanskrit ', merely on the score
that a few Dravidian words or echoes of Dravidian syntax might be
traced in it. I t would surely be better to retain the older use of
the term, and confine the description Buddhist Hybrid Sanskrit to
those texts or portions of texts which are, in fact, hybridized in
grammar ; and to distinguish the other texts simply as ' Buddhist
Sanskrit '.
Edgerton has stressed particularly the unity of tradition
running throughout the texts, and it is indeed important that this
should be clearly understood. But it is equally important,
particularly for an editor, to realize also the clear differences
between one Buddhist style and another. From one point of view the
history of Buddhist Sanskrit might almost be said to be a study of
the fluctua- tions of the degree of badness of the Sanskrit ; but
it is not all bad in the same way, nor for the same reasons.
The main outlines of the story appear to be as follows. The
earliest Buddhist Sanskrit authors (or compilers) had inherited a
considerable literature of canonical and semi-canonical texts
which, we may suppose, had been handed down chiefly by oral
tradition. I t has sometimes been said that this ' proto-canonical
' material must have been in some Prakrit dialect, and that the
Sanskrit and Pali versions of the canon represent two different
translations of the early canon with, of course, certain
modifications of the actual matter according to school. But there
seems to be no compelling reason for postulating a single Prakrit
dialect as the ' original ' language ; and it seems much more
likely that the texts were handed down in diverging ways in
different communities. If this is so, then the Pali might be held
to be a local crystalliza-
-
tion of a relatively fluid tradition, rather than a translation
as we would normally understand the term. The Sanskrit version,
however, is rather a different matter. The rendering of the
traditional material into Sanskrit would demand a much more
positive intention. So far as concerns the Sarv5stiv5din canon a t
least, there is no room to doubt that the authors fully intended to
write Sanskrit, and they would have been surprised at the
suggestion that they were writing in a language essentially
Prakritic in nature, since their whole effort was to present their
doctrine in the language of learning and prestige. The same desire
must surely underly Sanskritizing in the other texts also ; and if
the result in some places would have evoked the Brahman's derision,
the authors themselves were doubtless satisfied that they had
achieved something. The fact that they fell short of the classical
standard in
administratorHighlight
administratorHighlight
administratorHighlight
administratorHighlight
-
THE LANGUAGE OF THE BUDDHIST SANSKRIT TEXTS 369
varying degrees may on occasion be attributed to a simple
inability to write correct Sanskrit, since not all the monks had
had the benefit of a Brahmanical education. But it is significant
that the early Sanskrit Buddhist philosophical and literary
works-in effect, the non-canonical writings which are ascribed to
individual authors-are almost entirely in classical Sanskrit ; and
it is therefore likely that the chief reason for imperfect Sanskrit
in the early days was the resistance of the material itself. This
resistance has often been attributed largely to the exigencies of
the metre ; but probably a much more important factor was the
hieratic character of the texts, which the Sanskritizers would be
concerned to alter as little as possible. I t is understandable
that the verses would in this way be more resistant than the prose
; and that in the prose itself word-order, syntax, and vocabulary
would more readily persist than phonology and morphology. I t is
important to observe, however, that Middle Indian words and turns
of phrase, when retained in a Sanskritized version of older
material, would naturally tend to be accepted by later writers as
legitimate for use also in original compositions ; and where no
parallel in Pali has survived i t is usually impossible to say
whether a given verse was actually composed in Hybrid Sanskrit or
was transposed from a genuine Prakrit original.
Within the hybrid texts a number of fairly distinct styles can
be discerned. I quote here specimens of two extreme and quite
different varieties.
(a) tena ca yiithapatinii ye anye Bqapiyanti te pi na icchanti
gantuq, nasmiikam osaro, amukiiye mrgiye osaro, sB gacchatii ti.
sii ahaq tehi na mucyiimi osariito, vucyiimi gacchiihi tava osaro
ti, tad icchiimi mrgariijena ato anyaq mrgaq visarjamiinaq, yaq
velaq ahaq prasiitii bhavigyami tat0 gamigyiimi. so mrgarBjB mrgim
aha: tiiva mii bhiiyiihi, anyaq visarjayigyaq. tena mrgariijena
iiqapako mrgo iiqatto, ito yiithiito yasya mrgasya osaro t a q
iiqapehi, etaye mrgiye may%abhayaq dinnaq.
(iWahnvustu, i. 36%3 ; Edgerton's Reader, p. 3.) (b) ye jina
piirvaka ye ca bhavanti
ye ca dhriyanti dadadidi loke tepa jiniina karomi praqiiman]
te jina sari-a ahaq prastavipye (Xuvur?~ubh~sottu~~a-sutra,ed.
Nobel, p. 45.)
The first of these is hardly to be called Sanskrit a t all.
Apart from a thin veneer of Sanskrit spelling, it is typically
Prakrit, not only in many of its word-forms and inflexions, but
also in the stiff, awkward style characteristic of a good deal of
Prakrit prose. Indeed, we may reasonably surmise that it is
passages of this sort which underly the persistent tradition that
the Xahiisiinghikas used Prakrit, in contrast to the
Sarviistiviidins who employed Sanskrit.l The second example is
typical of stotra-verses in the dodhuka-metre, and contrasts
sharply with the first in the feeling of ease and flow in its
language. This admittedly
For a full discussion of this matter, see Lin Li-kouang,
L'aide-mimoire de la craie loi, 1949, pp. 176 ff.
administratorHighlight
administratorHighlight
administratorHighlight
-
370 J. BROUGH
may be due largely to the metre itself. But whereas the former
passage is beyond doubt composed in Prakrit, a great deal of the
stotras and similar verses in stronglyrhythmic metres may well have
been composed in what to the authors was essentially Sanskrit, with
the admission (by ' poetic licence ') of certain well-defined
abnormalities.
The less Prakritic portions of the iMahEvastu gradually tend
towards the style found commonly in the old prose Avadanas and the
canonical works generally, of which the following is a typical
specimen :-
(c) tena khalu samayena gandhamadane parvate raudrikgo nama
brahmaqag prativasati sma, indrajalavidhijnag. airaugid raudrikgo
brihmaqo bhadraiillyiq rijadhanyaq candraprabho nama r i j i
sarvaqdado 'smity atmanan) pratijanite. yan nv ahaq gatva iiro
yaceyam iti. tasyaitad abhavat : yadi tavat sarvaqdado bhavigyati,
mama diro dasyaty, api tu dugkaram etad asthanam anavakiiio yad
evam igtaq kantaq priyaq manipam uttamingaq parityakgyati yad uta
iirgaq, nedaq sthiinan) vidyata iti viditvi gandhamadanat parvatad
avatirqag.
(DivyEvdZna, p. 320.) This style might almost be called Buddhist
Sanskrit par excellence. I t is in general tolerably correct in
grammar, though it shows the Prakritic part of its ancestry in some
frequently recurring turns of phrase, and a fair sprinkling of
Middle Indian vocabulary.
A further development in the same direction, coupled no doubt
with the benign influence of good poets such as Aivaghoga, led in
some places to the use of an ornate Sanskrit which, apart from its
subject-matter, shows very few distinctively Buddhist features. A
good example of this, which might be called the semi-klvya style,
can be seen in the first version of the story of the tigress in
chapter xviii of the SuvarqabhEsottama-sBtra, which commences
:-.
(d) divi bhuvi ca
visytavimalavipulavividhagurlagatakiraqo'pratihata-jninadardanabalapadkramo
bhagavan bhikgusahasraparivytab pancalegu janapadegu
janapadacirikaq caramaqo 'nyatamavanakhaq4am anupripto babhiiva. sa
tatra dadaria haritamydunilaildvalatatavividhakusuma-pratimaqditaq
pythivipradedaq, dygtvi ca bhagavin Byupmantam anandam amantrayate
sma : dobhano 'yam inanda pythivipradedab.
(Suvarqabhcisottama-s4tm, p. 202.) It must, of course, be
recognized that here, as elsewhere, various gradations of style may
be found. Thus, chapter xxii of the DivycivadEna (from which
example c above is taken) presents passages of ornate semi-kivya
prose mingled with, and merging into, typical Avadina-prose ; while
most of the last chapter of the same work (the story of
Maitrakanyaka) might not unfairly be described- if we discount the
mediocre ability of the poet-as real Sanskrit klvya.
In the medieval period a great deal of the distinctively
Prakritic inheritance tends to fade. Most of these later texts are
as yet imperfectly explored, and I can give here only a few
tentative hints concerning their language. Except for the
-
THE LBKGUAGE OF THE BUDDHIST SAKSKRIT TEXTS 371
Manjuiri-mda-kalpa, they lie outside the scope of Edgerton's
grammar ; and of the language of this text he remarks that it seems
' bizarre, even for BHS ' (10. 4). There is no doubt that the text
as it appears in the edition is corrupt in many places ; but much
of it is reasonably typical of what might be called the didactic
style. This style is frequent in tantric works, though not confined
to them, and in its more extreme forms it may give the impression
that the authors were only semi-literate. The following specimens
illustrate a number of the commoner varieties of grammatical and
metrical divergences from classical Sanskrit.
( e ) tathaiva piijayet sarvailj samiidhitrayabhavanailj
ialidhanyaq ca Fat prasthaq Fagti dipaq prajviilayet.
tathaiva karayet piijiiq jiigareqa vini~kramaq
grahamiitykii(q) samabhyarcya yathoktaq grahasadhane.
daiame dviidaie vlhnau dviiviqiati dinegu va
niimakaraqaq prakartavyaq varqiinaq ca viiegatalj.
(Piipaparimocana, 17-19.) ( f ) priitar utthiiya iayaniit
snatvii caiva iuce jale
nibpraqake jale caiva sarinmahiisarodbhave
udghygya gatraq mantrajno mydgomayaciirqitailj
mantrapWaq tat0 krtva jalaq caukpaq sunirmalam
sniiyita japi yuktiitmii niitikalaq vilanghaye.
. . . . . . .
sugandhapugpais tatha iastu arghaq dattvii tu japinalj praqamya
iirasa buddhanaq tad5 tu iigyasambhavaq.
(Manjuirz-mcla-kalpa, pp. 97-8.) A very common feature of this
style, which naturally has no literary pretensions, is the frequent
occurrence of ellipsis and anacoluthon, though these do not
normally obscure the sense.
Rather different from these, and in general closer to classical
Sanskrit, is the language of the medieval verse Avadanas. I n its
better portions, in fact, it is hardly to be distinguished from
normal medieval narrative ilokas ; but in its less good parts,
occasional blunders appear which are not likely to be found in
Brahmanical works. The author of the following passage clearly
demonstrates by his verse-fillers and his jejune and awkward short
sentences the difficulty he experienced in composing in
Sanskrit.
(g) dadaria bhiipatir jirqa-praqaliq margake 'tha salj :
praviihitaq na piiniyaq, tad-darianena vahitaq.
' prak na praviihitaq, kena idaniq tu praviihitam ? '
aicaryeti sthite bhiipa, iikiiiiid enam iiciviin :
' puqyavaqs tvaq mahiiriija, tvat-prabhiiviit pravahitam.
bhagnabhiita praqiiliyaq, niinam abhyantare jalam.3
tvaq tu dharmatanur, evaq jlnihi kila bhiipate '.
(DvEviyiaty-avadEna,vi.) Ed. sanirmalam. i.e. ' (a voice) spoke
to him from the sky '. MSS.abhyantare yatab.
-
372 J. BROUGH
A still more striking lack of ability to control the language is
found in a later chapter of the same work (though presumably by a
different author), where the benefits of offering various kinds of
flowers to the Buddha is described :-
(h) brahmahatyii-iataq pipaq iatajanma-krtini vai
rohaqaq mighya-pugpiiqiq iamayati na saqiayab.
kokil ik~aq prarohante janaviin dhanavin bhavet
vidyiivatiiq kule jiitab sarvalokaib prapfijyate.
rohaqiic campakaq pugpaq nari ye iraddhayi kila kirtiiabdai ca
lokegu sarva-sampada labhyate.
(DvEviqSaty-avadlna, xv.) Here again the sense is quite clear,
alid the individual words for the most part appear to be Sanskrit ;
but they would defy any attempt at syntactical analysis in orthodox
terms.
The h a 1 decrepitude of Buddhist Sanskrit is reached in a text
like the Aivaghoga-nandimukhEvad~na.I have consulted four
manuscripts of this work, and I find it impossible to follow even
the thread of the story without constant assistance from the Newari
translation. Xaking every allowance for scribal corruption, which
is probably considerable, it would still appear that the author
wrote in a style reminiscent of a schoolboy's Latin prose
composition. The text of the following short extract is reasonably
certain, apart from the obelized word. (The Newari version has for
this ' looked around ', so that some form of saqlak~ayatiis
needed.)
(i)nandimukho aivaghogo devim iijnii iirasi nidhiya
matya-maqdalaq gatau, gatvi ca tat0 matya-maqdalaq carituq t
saqrak~ari t ra matya- maqdale. niinivicitropakiiranaq dptii l
abhiita jitau. aho matya-maqdale ramaqiyaq, kathanj caritavyaq
jniitavyaq. caqdikii-sthiine sthitvii vicitraq stri-riipaq
dhirayimi. gitaq karoti cintya, surabhi-manojna-gho~aq siikgmeqa
gitaq karoti sma. The foregoing account of Buddhist Sanskrit styles
is necessarily sketchy,
and makes no claim to be either final or exhaustive ; and it
must be recognized that any classification of the material in this
way is a mere convenience, a frame- work which we construct, within
which we can organize our thinking on the subject. Further study of
the texts will make possible a more detailed account. But it is
most important that anyone who undertakes to edit a Buddhist
Sanskrit text should be aware that there are such different styles,
and that features which are regular and common in one may be quite
unknown in another. A good deal of mischief can be done by an
editor who is not sufficiently conscious of these differences, the
most likely pitfall being the introduction of typical ' hybrid '
forms by way of emendation into passages which are either in
i.e. dl@tvB. i.e. adbhuta ' astonished '. One manuscript has, in
fact, atbhata.
administratorHighlight
-
373 THE LAKGUAGE OF THE BUDDHIST S,LWSKRIT TEXTS
real Sanskrit or in semi-kiivya style. An example of this may be
seen in the Suvarqabhiisottama-stltra, p. 210, where Eobel prints
:-
krpslkaruqasamudgatiiryasattvii divi bhuvi ceha ca labhyante
svadehaq : Qataia iha karonti nirvikiiraq muditamansllj
parajivitiirtham.
Here labhyante, karonti, and parajit~ithrtham are emendations of
the editor, who is thus willing to attribute to the author the
following lapses from grace : (1)labhyante, a passive verb used in
an active sense ; (2) karonti ; (3) mzcdita-manab, a singular
adjective with the plural subject ; (4) an incomplete sentence ('
for the sake of the lives of others they do . . . '-what ? Unless
perchance liaronti was taken to be an intransitive verb ; or
-artham thought to be its object, which is not only impossible in
itself, but is contradicted by the Tibetan phyir upon which the
emendation is based) ; and if, as seems most probable, the metre of
the stanza is puspitigrz, metrical irregularities in (5) kypii- ;
(6) labhyante ; and ( 7 ) the metre left two syllables short in
parajTzlitzrtham. Xow tllis stanza comes in the middle of a section
of the text in the semi-kiivya style, and was clearly intended by
the author to be in good, correct Sanskrit. I t is true that, in
order to accommodate some of the more elaborate metres, he allows
hiinself an occasional licence, for example, p. 211 dhyhnzdibhi
guqaifi (for -bl(ir), p. 215 bhratrqii (for bhratr6)-unless,
indeed, more deep-seated corruption underlies some of these. But an
accumulation of seven faults in a single stanza, without even good
sense resulting, is quite incredible. The editor remarks in a
footnote, ' Die \lTorte sind stark verderbt und unsicher ', and his
emendations might almost seem designed to ensure that they doubly
earn their obelus. The following readings (neglecting trifles) are
presented by the manuscripts : krpalcaru@- G, krp6karu~a- P l ;
karota G, karoti P ; parijiaites marair G (-s with virhma, and
marair dislocated in the manuscript), parajivite darire P. It may
be observed also that the Tibetan version confirms the plural
-sattviifi (sems can dug), while rnuditamanab demands a singular in
the second half of the verse. If then we may allow that the poet
might have written lirpa- for krpii- for the sake of his metre, I
would suggest the following for consideration :-
krpakaruqasamudgatiiryasattvii, divi bhuvi ceha ca labhyate
svadeham : Qataia iha karotu nirvikiiraq muditamaniib
parajivitopakiiram.
' Koble beings are born of pity and compassion, and an om-n-body
is obtained either in heaven or here on earth : (therefore) here on
earth, with joyful mind, one should in a hundredfold ways
unremittingly do that which is of service to the lives of others.'
This, I think, does less violence than the edition to the
1 P = the consensus of the interpolated group of
manuscripts.
-
374 J. BROUGH
manuscript tradition and to the language,' and it is metrically
satisfactory ; and while no absolute certainty is claimed for it,
it is a t least in keeping with the generally Sanskritic nature of
the story as a whole.
Equally with the editor, the interpreter and grammarian must
guard against seeing hybrid forms in styles where they are prirna
facie improbable. For example, Divyavadci?~a405, printed in the
edition as :-
Mauryab sabhrtyab sajanab sapaurab
sulabdhaliibharthasuyaptayajna4) : yasyedrial~ siidhujane prasiidab
kale tathotsahi k r taq ca danam.
(We may note in passing that the second pada should be emended
to sulabdhalcibhad ca suyastayajnab, and that in the third i d y bb
should be corrected to idyie, since the nominative absurdly makes
the king appear to be praising himself, whereas the whole context
shows that it is the monks who are the objects of praise.) In the
fourth piida, Edgerton (8. 60) understands utsahi to be a metrical
shortening for utsahe, ' co-ordinate with kale '. This seems to me
to be entirely ruled out, not only by the difficulty of construing
the verse in this way, but also by the fact that the verses in this
section are in kavya Sanskrit, with very few lapses. In the present
verse yagta for the correct i s ta is much easier to accept than -i
for -e. I suggest, therefore, that we should take utsahi-kytatg as
a compound. The sense of the stanza would then be that the Maurya
king (the speaker) really is a king, ' since his liberality is
bestom-ed on such holy persons, and since the gift, made by one
full of religious enthusiasm, has come so appropriately a t the
right time '.
These examples, together with the specimens of styles, will
suffice to indicate the attitude which I believe editors ought to
maintain towards their material. It is now no longer possible to '
correct ' indiscriminately the language of the Buddhist Sanskrit
texts so as to bring it into as close agreement as possible with
classical Sanskrit ; and to Edgerton, more than to any other
The translation given implies a ' hyper-sandhi ' in the first
pcida, for -samudga~ Bryasattocij. (This could be avoided if we
understood the whole of the pcida as a compound in the vocative
plural, ' 0 noble beings, born of pity, etc.' But such a vocative
is out of keeping with the context, and would also imply that the
second ca is merely a verse-filler.) The Tibetan translation urould
seem to support the view that the ciryasattocij are not the subject
of the second half of the verse, but are merely held up as a model.
This a t least seems a possible interpretation of the two small
additions to the word-for-word rendering : ' Considering that (seam
ste) noble beings are produced, etc. : in like manner (hthun par)
for the sake of the lives of others (I)shall show pity '. (The
subject of course is indeterminate.) Corresponding to the
conjectured karotu, upak&ram, the Tibetan has only brtse, v. 1.
rtse, of which the former is taken by the Kalmuck version (eneriku,
have pity, Altan Gerel, ed. E. Haenisch, p. 106.12), and the latter
by the east Nongollan (I.J. Schmidt, Grammatik der ,Mongolischen
Sprache, p. 166, erfreuen-though the normal sense of rtse in the
dictionaries is ' play, sport ', Skt. krig-.) The Chinese versions
given no ass~stance, since Dharmak~ema omits the verse entirely,
and I Ching has in its place an entirely different passage (Taish6
Tripitaka, xvi, p. 354b, and p. 451b, c). I am grateful to my
colleague Professor W. Simon for his assistance in comparing all
these versions with the Sanskrit text.
The other formal possibility, that idr8nj might be taken as
agenitive with yasya, is an-kurard and a t best yields a tame
sense, and seems to me most unlikely.
-
375 THE LAXGUAGE O F THE BUDDHIST SANSKRIT TEXTS
single scholar, is due the credit for this advance in our
understanding. But it is all too easy, as these two examples
shorn-, to fly to the other extreme, and work on the implicit
assumption that ' anything is possible in Buddhist Sanskrit ' ; and
an editor must try to adopt a madhyamii pratipad. The immediate
task for the future is the closer delineation of the various forms
and styles of the Buddhist writings in Sanskrit, and a detailed
grammatical analysis of each type. Those who undertake this task
will find in Edgerton's Grammar and Dictionary an invaluable guide
to a very large part of the field, and an indispensable work of
reference not likely soon to be superseded.
YOL. XVI. PART 2.
administratorHighlight