Open Collections

Digital Himalaya Journals

Does Tibetan Hermeneutics Throw Any Light on Sandhabhasa Broido, Michael M. 1982

Item Metadata


JSON: dhimjournal-1.0364974.json
JSON-LD: dhimjournal-1.0364974-ld.json
RDF/XML (Pretty): dhimjournal-1.0364974-rdf.xml
RDF/JSON: dhimjournal-1.0364974-rdf.json
Turtle: dhimjournal-1.0364974-turtle.txt
N-Triples: dhimjournal-1.0364974-rdf-ntriples.txt
Original Record: dhimjournal-1.0364974-source.json
Full Text

Full Text

Michael Broido
The importance of hermeneutics as "a philosophical discipline of rational
interpretation of a traditional canon of Sacred Scriptures" within Buddhism has been pointed out by Thurman.1 Buddhist texts frequently contain
detailed and sophisticated arguments about the interpretation of texts (even
themselves). This hermeneutical self-consciousness applies also to the scriptures (both sutras and tantras) traditionally thought of as uttered by the
Buddha himself, regarded as a historical person. Accordingly, there are
two different senses in which a Western work which is to count as a worthwhile interpretation of Buddhist texts may be concerned with hermeneutics. First, the Western interpreter must be aware of himself as an interpreter, as bringing hermeneutic techniques to bear on his materials; in this
respect he will of course be in a position similar to that of any other scholar
dealing with religious texts. But secondly, he has to take into account the
fact that, on the whole, the texts themselves were written with a certain
degree of hermeneutical self-consciousness, in the expectation that they
would be subject to, or would even require, interpretation. As Thurman
points out, this was the case with the earliest Buddhist texts, most of which
have a fairly straightforward "literal" sense. It will be all the more true of
later texts such as the tantras, of which many passages do not seem
"literally" to mean very much at all, while others contain admonitions,
such as that to kill one's fellow-creatures, which seem to run counter to
everything which Buddhism has otherwise (and on good grounds) been
held to stand for. Now in India and Tibet the tantras were taken seriously
as Buddhist religious documents, and so it is not surprising that a substantial corpus of interpretative literature grew up alongside them.
From this alone it will not follow that, when interpreting the tantras, we
must follow Buddhist hermeneutical tradition. But it does seem plausible
that an understanding of Buddhist hermeneutics will help us; and this point
seems especially important in view of the fact that most Buddhist tantric
texts now available are in Tibetan, either translated from the Sanskrit or
written by natives. For it would be remarkable if, in so translating and
* The form of this paper has been greatly improved as a result of suggestions made
by Dr. Nathan Katz. On specific points, my indebtedness to Mr. Alexis Sanderson
and to Dr. Brian Loar have been recorded in the text and notes. Conversations with
Mr. Gorden Sundholm (on the theory of meaning) and Mr. Edward Henning (on
some of the Tibetan materials used) have also influenced what is written here in
various places. My grateful thanks are due to all these people.
writing, the Tibetans were not influenced by their own hermeneutical
views. Accordingly, if we wish to use these Tibetan materials in our search
for an understanding of the tantras as they were seen either in India or in
Tibet, we should try to understand the hermeneutical views which the
Tibetans displayed in them.
Both in India2 and Tibet3, attempts were made to systematize the explanatory methods used in the voluminous commentatorial literature. The
methods of explanation themselves are called bshad-thabs" in Tibetan, and
the systematic treatment of these methods was considerably developed in
Tibet by such distinguished writers as Bsod-nams rtse-mo (1142-82), Bu-
ston (1290-1364), Btsong-kha-pa (1357-1419), and Padma dkar-po
The tantric form of Buddhism is often called Vajrayana (in contrast to
the Hfnayana and Mahayana, often lumped together and called
Laksanayana). The role of yana as a hermeneutic strategy has been
discussed recently by Katz (see note 1), and need not concern us here.
Now, among students of the Vajrayana there is a long-standing
controversy over a group of Sanskrit terms such as sandha-bhasa,
sandhya-bhasa, sandhya-bhasita, etc. Do these Sanskrit terms mean
something like "twilight language"5"7 or "ambiguous use of language"8? Or
something like "esoteric meaning"' or "secret language"10? Or something
like "intentional speech"11'12 or "intentional language"13?
Looking at a wide range of primary sources, I find this lexical view of the
matter rather artificial. The words have been used in a variety of ways, in
many different contexts, and at different historical periods. Each of these
translations tells us something about the way the words were used, but
none tells us more than a small part. So we need to disentangle the different
strands of usage; and when we do this, we find that, in spite of the lexical
similarity, our group of words is much more heterogeneous than has sometimes been supposed. Because of the complexity of the issues raised and the
need to examine many sources, I will here consider one main problem: the
term sandhya-bhasa as used in the Hevajra-tantra.
As is well-known, this word is used there of a kind of "secret language"
or "code," for instance: "passion" stands for "wine," "strength" stands for
"meat." Is there any sense in which this use of language is intentional! Why
call it twilight language? The central point of my paper is quite simple. It is
this: in discussing these questions, we may wish to take into account what
Indian and Tibetan commentaries on the Hevajra-tantra say about the
"code." If we choose to do this, we are committed to using, or at least
taking seriously, the principles of interpretation actually applied in those
commentaries. And this is possible only if we know what those principles
were. So we have two things to do. We have to identify the principles of
interpretation actually used, and we have to understand those principles.
The suggestion that Tibetan texts on interpreting the tantras (bshad-
thabs) may throw light onto the controversy over sandhya-bhasa and its
relatives has been made, somewhat implicitly, by Wayman. In those texts,
we find the technical term dgongs-bshad (explanation by 'intention); and
the Sanskrit for this is perhaps14 sandhyd-bhasitam. This last word seems
very similar to the word used in the Hevajra-tantra, viz. sandhya-bhasa;
and this word, one might think, is just the sort of thing that bshad-thabs is
there to explain.
In the phrase bshad-thabs, "thabs" means "method" or "technique" or
"means." The primary sense of the word bshad-pa is certainly "to explain,"
and yet there are places where it is appropriate to translate it by "to say" or
"to assert"; and it has even been translated by "to express."15 Let us see
why, both in "bshad-thabs" and in bshad-thabs, there are reasons for preferring "to explain" to "to express." The primary use of "to express' is one
where the (logical) subject is a sentence, and the object—what is expressed—is a thought, idea, question etc. A sentence (or its utterance) expresses a certain content. In the primary use of "to explain," the object is
still some thought, etc., but the subject is now the person who explains. A
sentence does not, as such, explain anything; and even a proposition
explains another proposition only in a secondary and derived sense. Confusion in the use of these two words in translation seems to occur mainly
when they appear in the passive voice, with the proposition (or whatever)
as the grammatical subject, for a proposition may be both expressed and
explained. But if the object-language sentence says that an account is given
of some sentence, proposition etc., then we need the verb "to explain."
This is the most common case in bshad-thabs; and "to express" is then inappropriate. In my view, Tibetan bshad-thabs texts signal this distinction
clearly by using the verb rjod-pa (vac-, abhidha-)i6 where "to express" is
the notion expressed.
Bshad-thabs, then, is concerned with the explanation, and more generally with the interpretation, of Vajrayana texts. Now hermeneutics is
perhaps the interpretation of texts, and the subject-matter of the study of
hermeneutics is perhaps the techniques which, in a particular religious
tradition, may have been used for the interpretation of its texts. Com-
monsense suggests that in a culture very different from ours, such as that of
Tibet, the hermeneutical methods may well have been very different from
ours. On the other hand, if for a moment we abstract away from the particular texts under investigation and the particular tradition they represent,
we are left with nothing more than the theory of interpretation in general.
And this is, broadly speaking, just what philosophers17 of language call the
theory of meaning.
This very general methodological consideration may be reinforced by
another which is more specific. If we broadly review18 the explanatory
methods (bshad-thabs) used in Tibet, we see that some of them are
linguistic in a rather natural sense of that word, that is, they turn on
analysis either of the sentences of the text itself or of the connection between those sentences and the purposes ascribed by the commentator to
their utterance. And now if we examine the linguistic methods in detail1*
we find that some of them do concern distinctions of just those sorts found
in the West in the theory of meaning and the theory of speech-acts. The
method of explanation by 'intention (dgongs-bshad) turns out to concern
just such distinctions. This considerably simplifies our task, for it means
that we have, in the theory of meaning, a source of relevant and carefully
worked out concepts expressed in the language of interpretation (here,
English, of course).
As we remarked before, in the Hevajra-tantra,, "sandhya-bhasa" is used
in referring to a kind of code, for instance20 "madana" (lit. passion) stands
for "madya" (lit. wine). In order to avoid taking sides in the controversy, I
shall not in this context translate "sandhya-bhasa," but shall instead use
"Hevajra code" to refer just to this particular code. (Similarly,
"Guhyasamaja code."21 Thus no attempt is made here to preserve sense,
but reference is carefully preserved. It is necessary to introduce some such
device now, because as Wayman has shown, the Pradipoddyotana ms.
uses, in some places, the almost identical form sandhya bhasa. Now: is the
use of this word in PPD the same as, or similar to, the use of the corresponding word in the Hevajra-tantra! Since this is one of the main questions
to be investigated in this paper, we need some clear way of signalling the
In order to find out just what the Hevajra code was, one may use commentaries on the tantra in two (related) ways. First, much of the commen-
tatorial material explains the tantra directly, without using the technical
language of bshad-thabs. Second, we can see what technical methods, what
bshad-thabs, were used in those commentaries. At the risk of tedium I shall
stress again that the code (sandhya-bhasa) is treated in both these cases as
something to be explained, and not as something which explains something
else. In contrast to all this, we have the explanatory or perhaps hermeneutical technique, certainly part of bshad-thabs, called "explanation by
'intention" (dgongs-bshad, perhaps sandhya-bhasitam). Among other
things, we have to find out whether the hermeneutical technique (sandhya-
bhasitam) was in fact used to explain the code (sandhya-bhasa). Bearing
this contrast in mind, we may set out the phases of this comparison, which
will be the heart of the paper, under the following seven headings:
1. The Hevajra code: sandhya-bhasa (dgongs-skad)
2. The hermeneutic technique: dgongs-bshad (sandhya-bhasitam?)
3. The  connection  between  sandhya-bhasa  and  sandhya-bhasitam:
evidence from Sanskrit sources
4. The connection between dgongs-skad and dgongs-bshad: evidence
from Tibetan sources
5. Proceeding from evidence based on Tibetan sources to conclusions
about the Sanskrit terms
6. Consequences for the controversy on sandhya-bhasa
7. Some conclusions for interpretation technique in English.
It will be necessary to go into some of these points at rather tedious length,
because of the many mistakes in the literature. These mistakes are largely
mistakes of interpretation, and surprisingly many of them (see section 7)
are assertions unsupported by argument or evidence. Accordingly, I shall
argue for my conclusions. The general structure of this argument may
perhaps be made clearer by the following summary of the topics to be discussed under these seven headings.
1. The Hevajra code: (sandhya-bhasa) (sometimes: dgongs-skad)
1.1: the codewords grouped
1.2: the Tibetan terms for sandhya-bhasa correlated with these groups
1.3: the uses of the code in the vajra-song and elsewhere call for explanation of the secret signs (choma, brda) found in HT I.vii
1.4: Padma dkar-po on secret signs: their unconventionality; their secrecy
and the reasons for it
1.5: Kong-sprul on the individual code-words (brda-skad-rnams)
1.6: Preliminary conclusions on sandhya-bhasa: the exact sense in which
the Tibetan commentaries say it is a code (independently of bshad-
thabs arguments)
2. The hermeneutic technique: dgongs-bshad (sometimes: sandhya-bhasita)
2.1: the phrase dgongs-pas bshad-pa: the importance of the instrumental
suffix (cf. sec.3)
2.2: dgongs-bshad in the Pradlpoddyotana: analysis of the definition
there, and its deficiency
2.3: Btsong-kha-pa on dgongs-bshad: this deficiency made up; an illustration
2.4: various interpretations of "dgongs-bshad," but their differences are
not too important
2.5: Kumara's comparisons of dgongs-bshad with other methods already
makes it clear why dgongs-bshad is irrelevant to dgongs-skad.
3. Sandhya-bhasa and sandhya-bhasita: evidence from Sanskrit sources
The question framed carefully. On the bshad-thabs (hermeneutic) side, the
exact form of the Sanskrit is not too important, nor do the Tibetan case-
endings tell us much about the Sanskrit (but see sec. 2.1). The Sanskrit
sources tell us that there is opposition, but not between what. On the
Hevajra side, the Sanskrit texts are more helpful, but they do not make up
the deficiency.
4. dgongs-skad and dgongs-bshad: evidence from Tibetan sources
Here the available definitions of "dgongs-bshad" are much fuller, and it is
evident that the Hevajra code does not fall within it. The hermeneutic
(bshad-thabs) techniques (of the code) which are used by Tibetan authors
are summarized; these techniques are incompatible with the use of dgongs-
5. What conclusions about the Sanskrit terms might we hope to draw from
the Tibetan texts? Obviously the main plank of such an argument, insofar
as it rests on bshad-thabs, has already collapsed; but the other necessary
steps are pretty doubtful too. Consider the following:
5(a) The Tibetan for sandhya-bhasa is dgongs-skad.
5(b) The Tibetan for sandhya-bhasita is dgongs-bshad.
5(c) In a certain Tibetan context, the terms dgongs-skad and dgongs-
bshad are used in certain clearly related ways.
5(d) Therefore,   in   the   corresponding   Sanskrit   context,   the   terms
sandhya-bhasa and sandhya-bhasita are used in similarly related
Not one of these four points 5(a) - 5(d) is even approximately right, (a) is
just factually wrong; the Sanskrit term mentioned in (b) is poorly established, and it is not easy to see how to make up the deficiency; in the
Hevajra context, (c) is hopeless (this was section 4); the Sanskrit context
mentioned in (d) does not seem to exist.
6. Results for the controversy on sandhya-bhasa
We are to consider the group of terms sandha-bhasa, sandhya-bhasa etc.
First, the assumption that this group is homogeneous is untenable. The
word sandhya-bhasita (etc.) does not belong with the others. Looking more
carefully, we obtain two groups of (perhaps) related terms, but no longer
with any lexical similarity. It seems possible that at a period earlier than
that of our bshad-thabs texts, these two groups of terms were not, in fact,
carefully distinguished. But to make use of this is to abandon any attempt
to make use of bshad-thabs. On the other hand, the Tibetan (later) texts
indicate what bshad-thabs methods were used of the code. But this information is available from the commentaries without much specific use of
bshad-thabs, which has now become irrelevant in a quite different way. If
there is any general conclusion, it is that the use of these terms must be
understood first in specific contexts. When this has been done, we may
perhaps be able to find a uniform interpretation. But there can be no a
priori ground for assuming this.
7. Some conclusions for interpretation technique in English.
We need a clear adequacy criterion for our interpretations. A criterion is
suggested, taken from the theory of meaning. This criterion enables us to
understand in a systematic way why some current interpretations of the
technical terms of bshad-thabs, even of such basic terms as sgra (sabda) and
don (artha), are so badly wrong.
1. The HEVAJRA CODE: SANDHYA-BHASA (sometimes: dgongs-skad)
1.1 The word sandhya-bhasa occurs nine times in HT Il.iii22; there are no
variations of spelling, other than those related to inflection.23 Though the
sandhya-bhasa words have been listed many times24 we must say something about their organization. Having regard to the Tibetan commenta-
torial tradition, it is convenient to group them into three groups as follows:
A. vv.56a-60a. Here a typical example is the first Pada:
madanam madyam balam mamsam. ..
/ma da na chang ba la sha/
This means: "madana" (passion) stands for "madya"" (wine26);
"bala" (strength) stands for mamsa" (flesh, meat).
B. v.60bcd. We have:
dvihdriyayogam kundurum /
vajram bolakam khyatam padma kakkolakam matam //
/dbang-po gnyis sbyor kun-du-rum/
/rdo-rje bo-la zhes bshad-de//padma kakkola zhes-zer/
This means: "kundurum" stands for "the union of two organs;" "bola"
stands for "vajra," "kakkola" for "padma." (Though of course "vajra"
and "padma" are used as euphemisms, the commentators also frequently take them in other senses.)
C. vv.62-63, which concern the five Buddha-families. Typically:
Dombi vajrakulr khyata Nati" padmakuli tatha/
/g.yung-mo rdo-rje'i rigs-su bshad//gar-ma de-bzhin padma'i rigs/
This means: "Dombi" stands for the vajra-family, "Nati" stands for the
All three groups of usages A - C are clearly called "sandhya-bhasa".
Groups A and B are used as a code, e.g. in the vajra-song27. In the case of
group C this is not obvious in the tantra itself. But as we shall see, the
Tibetan commentaries describe the usages of all three groups explicitly in
ways which strongly invite translation by "code". Further, groups B and C
are described similarly in the quoted passages by the use of the word
khyata (bshad). A literal translation of this word as here used might be
"explained as meaning," where "as meaning" is carried by the case-endings.
But these case-endings are different in groups B and C. This difference is
reflected in my translation in the absence of quotation-marks around
"vajra-family," "padma-family," as compared with their presence around
"vajra," "padma." For "Dombi" does not stand for "vajrakula," while
"bola" does stand for "vajra;" this difference of function is fairly clear in
the case-endings, both in Sanskrit and Tibetan.
Strictly speaking, a code is the substitution of one word for another
word, and is not a kind of naming. So strictly speaking, group C should
not be described by the word "code." As we will see, the Tibetan texts do
attempt to record this distinction by translating "sandhya-bhSsa" in two
different ways, but their usage is not uniform, and so with this caveat I
hope the reader will permit me to continue the use of "code" for all three
groups. Let us then see how the Tibetans did translate "sandhya-bhasa"
here, and what they did with the results.
1.2 The Tibetan for "sandhya-bhasa"
In HT Il,iii we find the following words:
dgongs-pa'i skad: vv. 1, 53, 54, 55
gsang-ba'i skad: vv. 61, 64, 65, 66, colophon.
These phrases contain the morphological feature -pa'i, -ba'i, which is often
described syntactically as a genitive case-ending, but which in cases like
this has the function of indicating that the preceding words are attributive
adjectives. Thus gsang-ba'i skad does mean, absolutely literally, "secret
language;" and if dgongs-pa means "intention," dgongs-pa'i skad means,
absolutely literally, "intentional language." (There can, of course, be no
question of "gsang-ba" or "dgongs-pa" meaning "twilight.") Clearly, then,
the Tibetans did not see the ending in sandhya as instrumental in this context. So if we find instrumental endings used in translating
sandhya-bhasita, we may feel that this use is deliberate.
Our two Tibetan words are directly associated with the three groups A
-C in the tantra thus:
A: three occurrences of "dgongs-pa'i skad," vv. 53, 54, 55
B: none
C: one occurrence of "gsang-ba'i skad," v. 61.
The remaining five occurrences ("dgongs-pa'i skad" once, v.l; "gsang-ba'i
skad" four times, vv. 64, 65, 66 and colophon) seem most plausibly taken
in connection with the passage as a whole, and not with any particular
portion of it.
Clearly, there is here a case for saying that "gsang-ba'i skad" refers to
these usages (the code) in general and possibly to group C, while "dgongs-
pa'i skad" refers to group A and probably group B, and is a particular kind
of gsang-ba'i skad. This possibility is not only interesting in itself (as undermining the 1-1 connection between "sandhya" and "dgongs-pa"), but is also
part (only) of the case for translating "sandhya-bhasa," as used of the
Hevajra code, by "secret language," as Snellgrove has done; we shall see
that there are other grounds for taking this translation seriously.
This feature of the Tibetan translation of the Hevajra-tantra has been explained by commentators along (at least) two different lines. These different lines of explanation make use of the fact that, in the tantra, there is no
occurrence of "sandhya-bhasa" specifically associated with group B.
Bkra-shis rnam-rgyal makes use of this to associate groups B and C tog-
ther. His account emphasizes the purpose for which the locutions are used.
He discusses group A under the heading2' "language suitable for the crowd
of yogins," and this phrase corresponds, in a very loose sense, to dgongs-
skad. Groups B and C are taken together under the heading30 "language
suitable for the mandala-cakra," and this corresponds in the same loose
sense to gsang-skad.
In contrast, Kong-sprul31 takes the distinction in a linguistic sense. He
includes both groups A and B under dgongs-skad, which he describes as
"normally using the name of the effect to stand for the cause."32 Only
group C comes under gsang-skad. If we were going to rely solely on Kong-
sprul's version, then, we would have a stronger case for abandoning the
description of group C as "code." When we come to discuss Kong-sprul's
use of the term "brda-skad" (section 1.5), we will see this view reinforced.
Kong-sprul has thus abandoned all attempt to bring the use of sandhya-
bhasa even in this one context under one sense. But Bkra-shis rnam-rgyal's
version, by taking group B with group C, leaves the possibility still open to
say that both dgongs-skad and gsang-skad are some kind of code. I am not
asking the reader to make a choice between these two versions. They are
exhibited in order to show that the writings of well-known Tibetan authors
contain a variety of views which will have to be taken carefully into
account if our interpretation of "sandhya-bhasa" is in the end going to be
claimed to have the sanction of the Tibetan commentatorial tradition as a
1.3 Uses of the Hevajra code
The main use of the code is in the vajra-song, which has been discussed
by several Western authors33'34 and has received a great deal of attention in
Tibet.35 The code is also used in other passages of the tantra, such as that
on feasting.36
Now an outstanding feature of the vajra-song is that many of the
unusual words found in it are not in the sandhya-bhasa chapter. The first
line of the song runs37
Kollaire tthia bola Mummunire kakkola
In their version of the tantra, the Tibetans wisely did not translate this; nor
shall I. The two capitalized words are place-names of a sort, and do not
appear in HT Il.iii. But at least "Mummuni" appears in the list of "places"
in HT I.vii. Though the vocabulary of the tantra does thus suggest some
kind of connection between the three chapters I.vii, Il.iii and II.iv, we have
to appeal to the commentaries to find out what this connection was taken
to be.
The whole commentatorial tradition makes it clear that the Hevajra code
(sandhya-bhasa) is part of something more general, viz. "signs" (choma,
brda). These appear in HT I.vii; and the "places" of that chapter are
examples of such "signs." What we now have to do, then, is to consider the
relation between the "signs" of I.vii and the "code" of Il.iii-iv.
1.4 Padma dkar-po on signs (brda)
Padma dkar-po (1527-92), possibly the greatest Tibetan scholar of the
Vajrayana and certainly a celebrated mystic, wrote a study on the Hevajra-
tantra" which, without going into much detail, groups the various themes
of that rather disorderly work in a way which makes them much more
coherent than do most commentaries. Like many of his other works, it is
written in a cryptic, awkward style and assumes that the reader already has
a good knowledge of the subject. So some of his sparse observations may
usefully be filled in with the more detailed but less incisive comments of
Kong-sprul.3' With Padma dkar-po, we may wish to distinguish between
the following matters:
The notion of a sign (brda) in general
The notion and purposes of signs (brda) in the Hevajra-tantra
Particular types of sign (brda) used in the Hevajra-tantra:
Bodily signs (lus-kyi brda): HT I.vii
Speech-signs (ngag-gi brda): HT Il.iii
Normally in Tibetan texts translated from the Sanskrit, brda translates
sahketa, and in Buddhist Sanskrit this word basically means sz'gn. It is not
really necessary to specify that these signs are conventional, for no sign can
designate something unless its use is governed by rules.40 But the rules may
be well and generally known, or less well-known. In the Hevajra-tantra the
signs are called choma, and though this word is also translated by brda,
Padma dkar-po41 makes it clear that not any sign counts, in this context, as
The essence of brda is that communicative intention is indicated by
speech or gesture not following normal conventions.
This definition is analyzed in note 41; my interpretation is founded on the
points made there about Padma dkar-po's vocabulary. He continues with a
sham-etymology (nges-tshig, nirukta) of brda:"
(Choma) is (here) said to be brda, because it is like the language of the
barbarians (kla-klo), by means of which the yogins recognise each
other but cannot be recognized by outsiders.
Thus choma (signs) are related to the secrecy of the tradition; and Padma
dkar-po even says that this is the point (dgos-pa) of brda. Such facts give
support to the translation43 of brda by "secret signs," and I shall so translate
it. Later we will see that "brda" is used (via the phrase brda-skad, lit. sign-
language) to explain the terms dgongs-skad and gsang-skad. This view
about "brda" (by no means confined41 to Padma dkar-po) lends further
weight to the translation of "gsang-skad" by "secret language."
Secret signs (choma, brda) are associated with vows and secrecy in two
slightly different ways. First, they help the yogin to recognize those
initiates with whom he may have congress, in a situation (a gana-cakra)
where uninitiated people may be present.46 Second, the use of secret signs
helps him preserve his vows and avoids the wrath of the guardian deities.4'
(Of course these points are related.)
The secret gestures (lus-kyi brda) are dealt with in detail in HT I.vii.
However, only part of the secret speech-signs are dealt with there, namely
the names of the "places" (ptthas etc.). The general account of how the
secret speech-signs are used is given with the explanation of the rest of the
individual speech-signs, i.e. the sandhya-bhasa words, in HT Il.iii.48 The
effect of all this is that the code-words do fall under the explanations of
Il.iii, but that these explanations have to be seen in the light of the more
general notion of secret sign (choma, brda) in I.vii. And this is reflected in
Kong-sprul's terminology where, even in commenting on I.vii, he refers to
the language of secret signs in general by using the word dgongs-skad.
Now that we are clear what brda means in the present context, we can go
back to HT Il.iii. Here, the commentaries explain the individual words in a
way which really does make the idea of a code rather explicit. The word
used by Kong-sprul (and others) is brda-skad; let us see how this word is
1.5 Kong-sprul on code-words (brda-skad)
In the following passage, Kong-sprul4' sets out the connection between
code (dgongs-skad) and code-words (brda-skad):i0
That which is called dgongs-skad is a passage accompanied by 'intention (dgongs-pa-dang-bcas-pa). What kind of language (skad) is this?
It is the code (brda-skad) of the unsurpassed vows made in the yogihi1
tantras. This (code) is not known to those who definitely belong to the
types (rigs) of the sravaka and pratyekabuddha and who have entered
the Hlhayana. Nor is it known to those who, having entered the
Mahayana, (adopt) the other, incomplete divisions of the tantra, that
is, the kriya-tantra of those who desire to smile at each other, the
carya-tantra of those who desire to gaze at each other, the yoga-tantra
of those who desire to embrace and kiss each other, and the anut-
tarayogatantra51 of those who, as a result of their karman, desire to
engage in the union of the male and female organs. Even these four
classes of tantra do not say52 that the dgongs-skad is such-and-
such ....
Shortly after, Kong-sprul begins his explanation of the details53:
Having told his questioner Vajragarbha to listen with fixed and undis-
tracted mind, the Buddha sets out to instruct him on the items of the
code (dgongs-skad-rnams) in proper order. These symbol-items (brda
'di-dag-rnams) usually refer to the cause by the name of the effect54.
Thus "madana" means "intoxication" (myos-byed), and brings about
drunkenness (bzi-ba), so it is used to refer to beer (chang), since when
one drinks beer, one gets drunk.
There follow numerous explanations on this pattern. Now at least eleven of
these explanations55 all follow the following scheme:
/rus-pa'i rgyan mtshon-pa'i brda-skad ni/ ni-ram-shu'o/
This means: As for the brda-skad standing for "rus-pa'i rgyan" (bone-
ornament), it is "niramsu." It is hard to think of more convincing evidence
that at least56 here in Kong-sprul, "brda-skad" is being used with the sense
(and not merely the reference) of "code-word."
1.6 Preliminary conclusions on the Hevajra code (sandhya-bhasa)
We have seen that there have been serious attempts to make a distinction
in sense between "dgongs-skad" and "gsang-skad;" this is important, but let
us ignore it for the moment. The whole tradition brings the code (sandhya-
bhasa, i.e. both dgongs-skad and gsang-skad, if these are different), under
secret signs (choma, brda). Kong-sprul says that the individual terms,
sometimes called e.g. dgongs-skad-rnams" or gsang-skad-rnams, are all
"code-words"—brda-skad-rnams. And these conclusions may be reached
without any reliance on bshad-thabs materials, from Tibetan writings on
the Hevajra-tantra. Knowing independently, then, what the Hevajra-code
is, and knowing also its basic purpose, viz. secrecy and the guarding of
vows, we have a solid foundation for assessing and understanding what
bshad-thabs tells us about this code.
This discussion has made use of only a very small part of the immense
range of writings available on the Hevajra-tantra. It is of course perfectly
possible that other sources will express other views. There is no harm in
this; of course the results are relative to the sources on which they are
based. The writers on whom I mainly rely for the Hevajra-tantra, viz.
Bkra-shis rnam-rgyal, Padma dkar-po and Kong-sprul, also wrote on
bshad-thabs and had roughly compatible views in that area too. Again,
there will be other writers of whom this can be said. For instance, the Sa-
skya-pa school had a very vigorous tradition of commentary58 on the
Hevajra-tantra, and used a system of bshad-thabs called "six instructions"59
different from the "seven ornaments" 60 system favored by our Bka'-
brgyud-pa writers and also by the Dge-lugs-pas, such as Btsong-kha-pa.
Now, the term sandhya-bjasitam belongs to the "seven ornaments" and not
to the "six instructions." This is why we concentrate on Hevajra works
from those traditions which, in bshad-thabs, use the "seven ornaments."
(sandhya-bhasita ?)
2.1 The phrase "dgongs-pas bshad-pa"
It is, 1 fear, a tedious feature of this paper that so many Sanskrit and
Tibetan phrases appear without a straightforward English equivalent; but if
the words sandhi, sandha, sandhya, sandhyaya etc. possessed such equivalents, the controversy over "sandhya-bhasa" would probably never have
arisen. "Sandhi" etc. derive from the root sam + dha-, meaning to place
together, to associate. When rendered in Tibetan by "dgongs-pa," these
words usually carry the notions of referring to, meaning, intending, and
sometimes intending deceptively. The weight, as it were, of these factors
varies greatly. Since the use of all these wrods tends to carry the notion of
purposive action, purely as a slogan I have represented "dgongs-pa" by
'"intention." Then the literal representation of the full phrase dgongs-pas
bshad-pa is "explanation by 'intention," or "to explain by 'intention."61
This full phrase is used quite regularly in some important sources, including
JVS and D. The ending on dgongs-pa is always -pas in this context (never
-pa'i, -nas, etc.) The most natural way to take the ending -pas is as an
instrumental case-ending, and this suggests that 'intention (dgongs-pa) is
the means or instrument by which the explanation is made. This suggestion
is basically bright, and it is valuable, because, as we shall see in section 3,
the corresponding situation in the Sanskrit is far less clear. In any case, as
observed in section 1.2, we may be sure that the use of the ending -pas by
the Tibetans was the result of deliberate choice; for in the case of the
Hevajra code, sandhya-bhasa, the typographically identical word sandhya
was translated by dgongs-pa'i.
2.2. dgongs-bshad in the Pradipoddyotana
Steinkellner<2) has rightly drawn the attention of scholars to the importance of the Pradipoddyotana in the historical development of Vajrayana
hermeneutics (bshad-thabs). The verse in which dgongs-bshad is defined is
quoted in Sanskrit by Wayman'6-7' and runs thus62:
visista-ruci-sattvanam dharma-tattva-prakasanam /
viruddha-alapa-yogena yat tat sandhyaya-bhasitam, //
Some of the details of the Sanskrit will be discussed in section 3. For the
moment, it is enough to observe that the verse must mean something like
this: "Whatever is to be explained 'intentionally is spoken by means of
contradictory (viruddha) discourse, revealing the suchness of things to
beings desiring the best." The Tibetan translation63 is perfectly straightforward:
/mchog 'dod sems-can-rnams-kyi phyir//chos-kyi de-nyid rab-ston-pa/
/'gal-ba'i tshig-gi sbyor-ba yis//gsungs-pa gang yin dgongs-bshad-do/.
Bhasitam = bshad here means explained and not expressed, not only on
the grounds given in the Introduction, but because nothing can be expressed by means of a connection (yogena, sbyor-ba-yis). Viruddha = 'gal-
ba means "opposed" or "opposing" or "contradicting," etc.64 This much is
(or ought to be) obvious. The real problem is about the phrase viruddha-
alapa-yogena = 'gal-ba'i tshig-gi sbyor-ba-yis. The word alapa normally
means "discourse" all by itself; what then is alapa-yoga? It seems plausible
to suppose that the connection (yoga) is .with the content of the discourse;
the phrase then means exactly an opposition between discourse and
content, such as we find in metaphors and deceptive utterances. This conclusion has been reached elsewhere (see note 3) by analysis of Tibetan
bshad-thabs texts; but here of course it rests on the merely plausible
supposition about the connection. The Tibetan version of the verse yields,
by itself, nothing further, and I believe that this is the best that can be done
with the verse itself (without assistance, say, from commentaries). Let us
now see how this common-sense interpretation of the opposition
(viruddha) just offered squares up with the fuller and more detailed
account of Btsong-kha-pa.
2.3 Btsong-kha-pa on dgongs-bshad
In his commentary (rgya-cher bshad-pa, lit. "extended explanation") on
the Jhanavajrasamuccaya," Btsong-kha-pa takes his definition of dgongs-
pas bshad-pa from the Pradrpoddyotana because JVS does not provide
adequate definitions of the "six alternatives" (mtha'-drugbb), and he glosses
the verse thus:67
For the benefit of those sentient beings of sharp intellect who desire the
highest attainment (siddhi). whatever is spoken in discourse whose
conventions (sgra) oppose the intention (don) (of the utterance), is
explained 'intentionally.
Elsewhere68 it has been shown in great detail that the pair sgra (sabda) and
don (artha), which are contrasted throughout the topic of bshad-thabs, are
related broadly in the following way:
sgra: words, phrases, sentences; linguistic convention, linguistic
don: content of a saying, proposition asserted (etc.), purpose or
intention of a speech-act; reference, referent
There is no need to repeat those arguments here. But we may note that if
sgra meant simply "word(s)," the passage would immediately give
nonsense, for words without linguistic meaning cannot oppose anything.
Similarly if don meant "meaning," the passage would again give nonsense,
because there would be nothing left in sgra which could be opposed by
don. Because it is so important to avoid the temptation to translate
sgra/don by word/meaning in this context, I will try to support the general
arguments of ref.3 by giving some specific ones.
There can be no doubt that Btsong-kha-pa intended us to take the opposition Cgal-ba) of convention (sgra) and intention (don) seriously, since he
expressed himself in almost the same way in several other places.69
In order to illustrate this theme, Btsong-kha-pa also gives a rather
extended example. To follow it, we must note that the difference between
dgongs-bshad and its opposite dgongs-min is never exemplified by two
interpretations of one and the same passage (this is the province of drang-
don and nges-don70), but always by two different passages having roughly
the same purpose (don).71 Btsong-kha-pa's example contrasts, then, two
remarks. The first says72
The purpose of purifying the three poisons (snag-ba gsum7') is to show
the radiant light (od-gsal).
This remark is to be taken straightforwardly, it is dgongs-min. By contrast, the
No desire, no lack of desire, and nothing is seen in between
is said to repudiate (bkag-pa") the previous remark, inasmuch as if there is
neither desire nor non-desire etc., how can there be anything to purify7
Taken literally, the two remarks are in opposition; but since the second
(like the first) is explained as showing the radiant light, the two remarks
have the same purpose (as ascribed by the commentator). In the first, the
linguistic meaning is in harmony with that purpose, in the second, they are
in opposition. (Obviously my two uses of "opposition" are essentially
From the point of view of religious experience, the Zen-like flavour of
the remark "No desire, no non-desire, and nothing is seen in between" gives
us a valuable clue. For the purpose of these strange utterances, in which the
sense of the words contradicts the utterer's intention, is said in JVS76 and
elsewhere to be to demonstrate the ultimate (mthar-thug-pa) which, being a
wordless experience, cannot directly be conveyed by words.
Another famous example with this flavour is that of "killing living
beings."77 The normal buddha-intention (of compassion towards all beings)
is opposed to the literal sense of the words. That is all that is meant here by
In both cases it is obvious that the difference is not merely one of words
(as opposed to meaning). For if we abstract the meaning away from the
words, we are left with nothing but strings of marks on paper (or sequences
of sounds) which cannot by themselves account for any difference in
modes of interpretation.
2.4 dgongs-bshad: other interpretations
It has seemed worthwhile quoting and explaining Btsong-kha-pa on
dgongs-bshad because his work D has been used by various Western
authors.78 Bu-ston78 and Padma dkar-po80 both gloss the PPD verse by saying
that the normal sense of the words opposes worldly attitudes. But Bkra-shis
rnam-rgyal81 and Kong-sprul82 both agree broadly with Btsong-kha-pa.
These differences are important, but they do not affect the present arguments. Elsewhere83, Padma dkar-po has given a deeper analysis of dgongs-
bshad, not based on PPD, which implies some criticism of the earlier
writers. His arguments are partly derived from ideas in the non-tantric
literature (e.g. MSL); but I cannot go into this here84.
2.5 Kumara's comparison
In his PPD-based analysis in F, Padma dkar-po quotes a passage from the
Indian author Kumara, who compares dgongs-bshad with other explanatory methods from the "six alternatives":
Whereas dgongs-bshad rests on differences of sgra (words and linguistic  meaning),   drang-don  rests  on  differences  of  don  (purposes
Whereas dgongs-bshad opposes, sgra ji-bzhin ma-yin-pa is a matter of
unknown signs (brda).
Whereas dgongs-bshad opposes, sgra ji-bzhin does not oppose.
Already the second of these remarks tells us why dgongs-bshad is irrelevant
to the Hevajra code. For we know that this is just a matter of signs (brda)
and of sgra ji-bzhin ma-yin-pa. The language of secret signs (brda), of code
(brda-skad) does not oppose anything, either the purpose of utterance
(Btsong-kha-pa, Bkra-shis rnam-rgyal, Kong-sprul), or wordly usage (Bu-
ston, Padma dkar-po). It is unknown in the world, it conceals its purpose;
and these are just the province of sgra ji-bzhin ma-yin-pa.
What, then, if any, is the evidence from Sanskrit sources alone that the
two terms sandhya-bhasa, used of the Hevajra code, and sandhya-bhasita
(or its variants), used of a certain method of explaining texts, are similar or
connected? Here it may be accepted that the first term is indeed the name of
a kind of text or passage or use of words which is to be explained, while the
second is the name of a method of explaining texts, etc. This important
distinction is not what is now at issue. The question before us is this: are the
passages, for instance in the Hevajra-tantra, called sandhya-bhasa, in fact
explained, say in the commentaries on that tantra, by means of the explanatory technique called sandhya-bhasita! Let us see what the Sanskrit
sources offer us.
On the sandhya-bhasita (hermeneutical) side, the first thing to do is to
establish precise forms of the Sanskrit under examination; for although it
will turn out that this is not certain, unless we can limit the extent of the
uncertainty there will be no basis for discussion. Here, Wayman seems to
have done most of what is necessary in ref.6 (and it is a pity that this, the
most useful part of that paper, is omitted in his otherwise very similar
paper, ref.7). Wayman shows that in the Bihar ms. of PPD, the normal
form is sandhya bhasa (two words), but that in the critical verse defining
the term (quoted above) we have sandhyaya-bhasitam. Given the predominance of forms in sandhya-, it seems reasonable to modify this to
sandhya-bhasitam. (By contrast the inflection on the second part of the
compound has a clear function and must be left). The normal Tibetan
forms are dgongs-pas bshad-pa and dgongs-bshad. Let us review the arguments about the case-endings in these phrases. First, the Sanskrit. I shall let
"bhasa" stand for the different forms of the second part of the word, since
they will not need to be discussed. Since sandhya bhasa is uncompounded
while sandhyaya-bhasitam is a compound, we must consider both compound and uncompounded forms. The possibilities seem to be:
(a) sandhi in instr. + bhasa (two separate words)
(b) sandhya in nom. + bhasa (compound)
(c) sandhya as indeclinable participle + bhasa (compound)
(d) sandha + bhasa (compound)
Here, (d) takes into account the possibility that -dhy- is a Nepalese orthographic variation for -dh-. Other combinations seem implausible because
there is no way of accounting for the connection of the two words. Way-
man's treatment in this area is somewhat inexplicit, but it seems he wants to
assimilate these cases. His argument seems to be that the (a)-like form at
Pahcakrama 11.31 is glossed by Sri-LaksmI with an instrumental (in
Tibetan), and the standard Tibetan form in bshad-thabs also has the instrumental. So in the end it does not matter which of the forms (a) - (d) we
adopt, as far as morphology is concerned (his adoption of "twilight
language" presumably favours (b) on other grounds). So far so good. But
the Tibetan instrumental85 does not mean, as he takes it, in the manner of
(which would normally be tshul-gyis, acting as a postposition with the
genitive). It means by means of. Because of this, the SrF-Laksmi"example
does not help Wayman. Her phrase (ref.6, p.790) "dgongs-pas bshad-pas
bstan-pa'i don-dam-pa'i byang-chub sems" means "the paramartha-
bodhicitta which is taught by explanation by means of sandhi (etc.)." (I
don't at all think the Sanskrit in the Pahcakrama does mean this, but that is
another matter.) In the Hevajra-tantra, the situation with sandhya-bhasa is
also not favourable for Wayman's argument, for as we have already
pointed out, that phrase is there always translated by a form attributive
adjective + noun (or morphologically, by a genitive ending). In any case,
in Pahcakrama 11.31 we also have a genitive (dgongs-pa'i byang-chub).
These arguments revolving around the detailed forms of the Sanskrit
words, then, tell against the conclusion which Wayman has tried to draw
from them. But I too find I can get only rather wavering support from
them. So I will not rely on these arguments at all, but will return to the
form appearing in the critical verse, viz. sandhyaya-bhasitam, merely
accepting that Wayman has made a good case for modifying it to sandhya-
bhasitam; and I shall use this form as a cipher for the Sanskrit term in the
PPD which is now under examination, bearing in mind that the ms. contains also the form sandhya bhasa (uncompounded). On bhasa/bhasita the
Tibetan is no help, since bshad-pa can be noun, infinitive or past participle
("explanation," "to explain," "explained").
Let us then return to the PPD verse itself. All this seems to tell us (section
2.2) is that the words or the text or the discourse (alapa) is opposed
(viruddha) to something. To what7 The guess that we have opposition
between discourse and purpose or content is only plausible, not more.
On the sandhya-bhasa (code) side, the situation is not quite so frustrating. Common sense suggests the connection between HT I.vii (on
choma), Il.iii (on sandhyabhasa) and II.iv (the vajragiti), for it is obvious
that the vocabulary for the latter is drawn from the two former. And the
purpose of secret signs (choma) in general and of the Hevajra code
(sandhya-bhasa) in particular is made clear enough by HT I.vii.l and the
Yogaratnamala on it: the initiates use a barbarous (miliccha) form of
communication so that they will recognise each other and not be recognized by outsiders, etc. (YRM and also the Tibetan commentaries are surprisingly vague on whether the word "choma" itself is barbarous, or its
referent. However, Mr. Alexis Sanderson kindly tells me that "choma" is
Middle Indo-Aryan from Sanskrit "chadman," "disguise." If this is right,
then surely it is the referent, the secret signs themselves, that are
barbarous.) Now obviously there z's an opposition here. It is the opposition
between the natural sense of the words used (say) in the vajragtti and the
non-natural sense in which they were intended to be understood by the
yogins and yogfnis (a sense explained both in the tantra itself, say at
Il.iii.56 ff., and in YRM [on the vajragiti]). And this opposition is of course
just the opposition which we find in sgra ji-bzhin ma-yin-pa, and not that
in dgongs-bshad.
But this last step cannot be taken on the basis of the Sanskrit texts alone.
For in the last analysis, the deficiency already noted in the PPD definition
of dgongs-bshad does not allow us to make the distinction which is here at
issue, between an opposition within alapa (discourse) or an opposition
between alapa and something else.
We could argue from the first line of the Sanskrit verse: sandhya-bhasa is
for those who desire the highest. But bereft of Btsong-kha-pa's gloss, that
these people are those of sharp intellect (dbang-po rnon-po), this does not
tell us enough. For those the code can be learnt by anybody, however
stupid, such persons are not necessarily devoid of ambition.
So we conclude that the Sanskrit texts presently available are not
enough, at least on the bshad-thabs side. We must turn to Tibetan texts.
Enough has already been done to show that, when indigenous Tibetan
sources are taken into account, the connection between the Hevajra code
(sandhya-bhasa) and the similarly-named hermeneutic technique is very
problematic. Now I want to go further than this, and to show that the
Tibetan texts decisively repudiate such a connection.
The question now before us is this: are the passages in the Hevajra-tantra
called e.g. dgongs-skad in fact explained by means of the explanatory
method called dgongs-bshad? The answer is unequivocal: in the texts which
I have examined, the dgongs-skad passages are never explained by dgongs-
bshad. They are explained by other methods; and the general structure of
these explanatory methods (bshad-thabs) is such that, given the methods
actually in use, the use of dgongs-bshad is impossible, it is incompatible
with them.
First, the methods actually in use. The linguistic conventions are nonstandard. Even the Sa-skya-pa tradition, which does not use the category
sgra ji-bzhin ma-yin, makes this very clear. And all the Bka'-brgyud
authors whom I have quoted say quite explicitly, in reference to dgongs-
skad, gsang-skad and brda-skad, in the Hevajra context, that all of these
are sgra ji-bzhin ma-yin-pa86 (na-yatharuta, not according to normal
convention). And as regards the interpretation of passages using the codewords, all our authors agree87 that these passages possess both drang-don
and nges-don (neyartha and riitartha) interpretations. And as Padma dkar-
po has particularly emphasized (though we know it also from Kumara and
Btsong-kha-pa) these methods, the ones that are actually used, can be used
only if there is at least rough compatibility between the normal use of the
sentence and its use in the passage under examination, whereas dgongs-
bshad can be used only if there is incompatibility. Here, we clearly have
compatibility; for the choma (brda, secret signs) are used with the basic
intention of keeping the tradition secret; and this is precisely in accordance
with the use of a code.
Indeed, once we thus understand the difference between the two terms,
many arguments become available. Btsong-kha-pa says that dgongs-bshad
is for the intelligent; but it is obvious that the code can be used by anybody
who knows it, however stupid. J VS and PPD both say that dgongs-bshad is
closely connected with mthar-thug-don; but it is obvious that the purposes
(guarding vows etc.) of the code have nothing whatever to do with mthar-
thug-don (pace Kanha on HT Il.iii.1). One could go on indefinitely.
Let us abstract for a moment from the fact that we deal with lexically
similar pairs of terms: sandhya-bhasa/bhasita, dgongs-skad/bshad. We
can easily do this, for instance, by thinking of those many passages in the
Tibetan commentaries which use the terms gsang-skad and brda-skad.
None of these Hevajra passages so much as mentions dgongs-bshad. Why,
then, other than because of the lexical similarity, should it ever have
occurred to anybody that the terms might be related?
In the review we mentioned four stages of this argument. The main
point, 5(c), is hopeless, and need not be discussed further. Let us consider
the other points, however.
5(a) is simply wrong. Sandhya-bhasa, as used in HT Il.iii, corresponds to
the two Tibetan words dgongs-skad and gsang-skad. It is a matter of controversy whether these two words are synonymous or not, even relative to
this limited context. Further, if there is a single Tibetan term which is used
of sandhya-bhasa generally in HT Il.iii, that term is gsang-skad and not
dgongs-skad. What these two have in common is that they are certainly
varieties of brda-skad; but this word cannot support any comparison with
5(b) is precarious. Here it is the Sanskrit word sandhya-bhasa which is
poorly established. It is hard to ignore the occurrence of a whole range of
similar Sanskrit phrases in such sources as the Lahkavatara," the
Saddharma-pundarika69 and the Mahayanasutralahkara.,° If we are going
to take the Tibetan tradition of bshad-thabs seriously, we cannot ignore
these non-tantric texts, since Padma dkar-po's account of dgongs-bhsad,
the best one known to me, is largely based on them. But as we shall see, in
these texts there is no straightforward one-to-one relationship between
"dgongs-bshad" and "sandhya-bhasita" (or any other pair of similar terms).
In relation to these sources, the Pradipoddyotana (with its few occurrences
of our terms) does not carry very much weight. (This situation might
change, if for instance Sanskrit mss. of subcommentaries on PPD or of the
Sandhi-vyakarana were discovered.)
Similarly, the Tibetan term dgongs-bshad suffers competition. For a
number of important Tibetan authors abandoned it and used instead
"dgongs-pa-can."'1 This is not just a matter of synonymy; a change of
meaning is involved.
The inference 5(d) is also precarious. Suppose the facts under 5(a)-(c)
were all in favour of the suggested hypotheses, rather than against them.
What conclusion would follow? One could conclude only that the Tibetans
interpreted "sandhya-bhasa" and "sandhya-bhasita" in certain closely
related ways, the relation being that set out in the beginning of section 3. If
this were true (which it is not), it would be interesting. But nothing would
follow about how these words were understood or used in India.
Conclusions about their use in India might be drawn from bstan-'gyur texts
on bshad-thabs and on the Hevajra-tantra. Nobody has ever claimed to be
able to do this. My impression (no more!) is that dgongs-bshad, which is
important in the Guhyasamaja system, was not used in India in the Hevajra
system. If this is correct, the basis for the comparison, insofar as it related
to India, is non-existent. On the other hand, we have already seen how the
Hevajra-tantra itself claims that its sandhyabhasa cannot be understood by
those who practice the father-tantras (such as the Guhyasamaja). Some
weight must clearly be attached to this claim.
In my opinion, this line of argument can hope to succeed only if one
abandons Guhyasamaja-based bshad-thabs altogether, and looks at the
bshad-thabs actually used in Indian commentaries on the Hevajra-tantra.
The notion of dgongs-bshad, and the whole programme discussed in this
paper, then probably become irrelevant. Obviously, then, one will expect
quite different results. I shall sketch one such proposal, which has some
attractive features, in section 6.
Of the group of words sandha-bhasa etc., some, such as sandhya-bhasa,
are undoubtedly used (e.g. in HT) to refer to kinds of language. Some, such
as sandhya-bhasita or sandhya-bhasa (both as used in PPD), seem to be
names of methods of explaining tantra passages.
However, the method sometimes called sandhya-bhasita never seems to
be used to explain the language-form called sandhya-bhasa. So our group is
heterogeneous, in a rather strong sense of that word. Such heterogeneity in
our group may, indeed, extend further.
If further investigation is to be based on Tibetan texts on bshad-thabs,
then the heterogeneity just noted will force us to consider separately at least
the following groups of terms:
(a) sgra ji-bzhin ma-yin, gsang-skad, dgongs-skad, brda-skad, brda
(c) dgongs-bshad,   dgongs-pa-can,   Idem-por  dgongs-pa,   Idem-po'i
There will also be words whose uses straddle the two groups, principally
the word dgongs-pa itself.
Correspondingly, there will be at least two separate groups of Sanskrit
(a') na-yatharuta, sandhya-sa (in HT), sandhi-bhasa (in connection
with GST), choma, possibly sahketa (e.g. in HT)
(b') sandhya-bhasita and sandhya bhasa (both in PPD); abhisandhi
and abhipraya (both as interpreted by Padma dkar-po)
Again, there will be terms which straddle the two groups, such as certain
uses of sandhi, sandhaya. And obviously this whole line of discussion is
directly relevant only to the uses of these words in the tantras. Obviously
too the purely lexical motivation for the analysis has now disappeared. But
I cannot imagine any worthwhile conception of philology in which this
motivation is any but the most tentative possible suggestion of meaning. In
our division, the motivation will be the difference between words used in
accounts of the linguistic conventions governing some rather unusual ways
of using language (a, a'), and words used in partly 'intention-based accounts of some rather different ways in which language has, on certain occasions, been used (b,b').
Now this distinction has been taken seriously in bshad-thabs; for, as has
been shown in so many different ways, it is the basic distinction between
sgra ji-bzhin ma yin-pa and dgongs-bshad. But to say this is not to say that
the distinction was made or recognized or presupposed, or that it would be
useful to make it, in other contexts.
The present study is based mainly on Tibetan texts written with the distinctions of bshad-thabs in mind. No attempt has here been made to trace
the historical evolution of these distinctions.'3 But it seems at least plausible
that there was a period when the group of notions, just discussed as
separated in Tibetan accounts of bshad-thabs, was not so separated, and
when explanations of sandhya-bhasa and other terms in the tantras did not
separate these notions. Let me briefly sketch this line of treatment of the
problem, so different from that followed elsewhere in this paper.
In the Mahayanasutralahkara on XII.16-18'4 and Sthiramati95 on it, we
find a lengthy treatment of linguistic usages which are not literal. The key
terms here are abhisandhi and abhipraya, translated in the Tibetan text of
the Sthiramati commentary by dgongs-pa and Idem-po'i ngag respectively.
Sthiramati's numerous examples include many which in bshad-thabs would
clearly fall under either dgongs-bshad or sgra ji-bzhin ma-yin-pa. The latter
term is in fact used by Vasubandhu'6 and Sthiramati97 of the type of
abhipraya called arthantarabhipraya (don-gzhan-gyi Idem-po'i ngag). The
treatment is organized in a way which makes it clear that this case is not to
be thought of as something separate from the cases of abhisandhi (dgongs-
pa) which are discussed. On the other hand, the examples given make it
clear that the notion of arthantarabhipraya covers such cases as the parable
of the sands of the river Ganges, which in the Lahkavatara9" is also
described as ayatharuta (sgra ji-bzhin ma-yin, not according to the literal
sense) and upamamatra (merely by way of comparison). So here, the two
notions, later separated under dgongs-bshad and sgra ji-bzhin ma yin-pa,
seem not to be separated or distinguished systematically.
The only available Sanskrit commentary on the Hevajra-tantra is the
Yogaratnamala. On the very word sandhya-bhasa at II.iii.53, this work
sandhir abhiprayah : abhipraya-pradhanam bhasanam :
naksara-pradhanam ity arthah :
Thus in "sandhya-bhasa", Kanha is taking "sandhya" as "sandhi", and of
this latter word he says that it means "implicit", that it is speech which is
mainly implicit, and not explicit."
The resemblance with the MSL vocabulary is obvious. (One might have
expected na-yatharuta instead of na-aksara; but whereas "na-yatharuta"
plays only a small part in the MSL discussion, aksarartha100 is important in
YRM). The general tone of YRM is Yogacara, and Kanha must have been
thoroughly familiar with MSL and its literature. This line of discussion also
provides us with an explanation of the remark at YRM on H.iii.l:
abhisambodhi-bhasanam samdhyabhasanam.
Padma dkar-po's account101 of dgongs-pa-can depends heavily on MSL
or some similar source. His vocabulary and examples are very similar to
those of Sthiramati. However he appears to translate abhisandhi by "Idem-
por dgongs-pa" (not simply "dgongs-pa"), and abhipraya by "dgongs-pa"
(and not "Idem-po'i ngag"). These points and their implications will be
dealt with elsewhere (see note 84).
The fact that this line of treatment of the "problem" of sandhya-bhasa is
available and has been used in Tibet shows us again the need for caution in
the use of Tibetan texts: specifically, for restriction of our claims to the
context in which our analysis was made. And it is largely for this reason
that I do not want to add my voice to those urging the adoption of one of
the lexical views of "sandhabhasa" etc. mentioned in the Introduction.
If (as the O.E.D. suggests) philology is the study of language in general,
the methods of this paper have been in great part philological. At the same
time, the conclusions are very different from those of the few other articles
which have used bshad-thabs. It may be worth sketching the methodolog-
ical basis of these differences.
As far as the analysis of texts is concerned, we may distinguish (at least)
three kinds of philological activity. There is the structural analysis of
sentences in terms of morphological, grammatical and syntactic notions.
Then there is lexicography, the attempt to give an account of the linguistic
meanings of words and phrases. Third, there is the process of interpretation: the attempt to furnish an interpretative description of the text in the
language used for analysis.
This paper has been concerned with what is, on the face of it, a controversy in lexicography: what does "sandhya-bhasa" mean? My conclusions, though rejecting the particular terms in which the controversy has
been carried on, are still largely lexicographic. But lexicography proceeds
in relation to the other disciplines; for the evidence as to the linguistic
meanings of words can be found only in their use in utterances, including
texts. Now it seems fairly clear that evidence of how a word was in fact
used in a particular utterance can be founded only on some kind of putative
description of that utterance, together with some kind of evidence or argument that the utterer intended by means of his utterance to convey to his
audience a content specified in the description. The vagueness of this is
deliberate; what else could count as evidence on which to base lexicography?
To refine this account, we may ask: what is to count as an acceptable
interpretative description, in English, of a passage in a foreign language?
The notion of translation is not much help here; for that notion, if it is
capable of clear definition at all, remains on the level which we need to get
below, as it were. Recent work in the theory of meaning102"4 has however
provided us with a notion of interpretation which does not depend on the
slippery notions of meaning, sense, reference or translation and can indeed
be used to explain those notions, yet is also not too behaviouristic to be
useful for philology. Let me quote McDowell:105
For that systematic imposing of descriptions to be acceptable, it would
have to be the case that speakers' performances of the (linguistic)
actions thus ascribed to them were, for the most part, intelligible in
the light of propositional attitudes: their possession of which, in turn,
would have to be intelligible in the light of their behavior—including,
of course, their linguistic behavior—and their environment.
Weighty philosophical questions turn on whether this test is sufficient for
acceptability. But we need not beg those questions here: for here it will be
enough to use the test in the very much weaker form of a necessary condition.
Though in this paper I have not, on the whole, expressed my own conclusions by offering translations, for the purpose of comparison with other
writers it will be convenient to explain McDowell's test in the case where
the description is a translation, or at least near enough to speak of a translation of individual words.
Among the propositional attitudes in question are the beliefs held by the
native writer about what is done in his writing-act. Perhaps he will believe
that his sentence forms part of an argument, or that he illustrates or applies
it later in the passage, etc. Our translation of the sentence must relate to
our translation of its context in such a way as to make these beliefs intelligible. And this will impose constraints on our translation of the individual
words of that sentence. For example if we translate what is thought on
some ground or other to be an argument, so that we are ascribing to the
writer a belief that his writing expresses an argument, our choice of words
must make it appear at least intelligible that the writer also believed that his
argument led to the conclusion which (we claim) he believed that it did lead
to. To say this is not, of course, to say that the argument, as translated,
must actually lead to that conclusion. Similarly, if we translate what is
thought to be intended by our writer as an explanation of some locution or
argument, then it must appear intelligible, from our translation, that he so
thought; this is not to say that the translation has to constitute such an
explanation; and so forth.
Put in this modest form, the point is related to one familiar in
philosophy. W. V. O. Quine has said106
The maxim of translation underlying all this is that assertions
startlingly false on the face of them are likely to turn on hidden differences of language. . . . The common sense behind the maxim is that
one's interlocutor's silliness, beyond a certain point, is less likely than
bad translation
However, McDowell's intelligibility condition, even in the weak form in
which I am using it, provides a far clearer criterion than does Quine's
It is just this test of intelligibility which so often fails in the quoted
papers. Over and over again, the translations of the technical terms are
inconsistent with the claims made (or implied), under translation, about
the content of the passages in which those terms occur. The effect is that,
under translation, the natives authors' claims become unintelligible in the
sense just sketched. I shall list systematically some examples related to the
"six alternatives" (mtha'-drug107).
(a) The translation of sgra (sabda) by "word" or "words," and of don
(artha) by "meaning".108 In section 2.3 it was shown that each of these
separately makes a number of observations by Btsong-kha-pa on dgongs-
bshad unintelligible: it becomes unintelligible that he believed that his
explanation should have the explanatory power he ascribed to it, or that
his example should have the illustrative power he ascribed to it. Concerning don, the same applies to his description and illustration of the difference between drang-don (neyartha) and nges-don (nitartha) in the same
(b) The translation of viruddha Cgal-ba) by "ambiguous"110 makes
Btsong-kha-pa's claim in the same dgongs-bshad context that there is
repudiation Cgeg-pa) unintelligible (note 69); for ambiguous discourse cannot repudiate anything.
(c) Separately from point (a) on don, the translation of nges-don by
"evident meaning"111 and drang-don by "hinted meaning" make unintelligible Btsong-kha-pa's (perfect correct) claim that they are used in the
tantras of one and the same passage.112 They also make unintelligible his
illustration mentioned under (a), in which nges-don is not in any way more
"evident," nor drang-don more "hinted" (quite the opposite in fact). They
also make unintelligible those many cases where drang-don is based on a
completely explicit code, while precisely nges-don interpretations are
hidden (sbas-te) and, as it were, hinted. Such cases are common in the
Hevajra-tantra and have been mentioned earlier.113
Further, the term "explicit meaning" is incoherent. (What is said may be
explicit, not what is meant.) So the claim114 that nitartha means "explicit
meaning" is unintelligible (regardless of context) unless it is seen as a claim
that "nitartha" is incoherent. But no such claim is there made or could be
seriously entertained. In particular, there is no claim, nor could there be
one, that "nitartha" is a metaphor.
(d) The translation of sgra ji-bzhin ma-yin-pa (na-yatharuta) by "coined
term"115 makes unintelligible our Tibetan authors' repeated claims that the
sandhya-bhasa of the Hevajra-tantra is sgra ji-bzhin ma-yin. For the
sandhya-bhasa words are mostly just ordinary words, in no way
"coined"."6 Further, sgra ji-bzhin ma-yin can be (as there) combined either
with drang-don or with nges-don, but it is unintelligible how the meaning
of a coined term could be "hinted," and it is pointless to describe it as "evident" (let alone "explicit").
The phrase "non-literal word" is incoherent. The claim117 that na-
yatharuta means "non-literal (words)" is therefore unintelligible unless seen
as a claim that "na-yatharuta" is incoherent; but no such claim has ever
been made or could be seriously entertained, especially as regards the
Tibetan "sgra ji-bzhin ma-yin;" for given what we already know about
"sgra", this phrase translates literally as "not according to a convention"
and this is almost exactly right. It is useless to claim that "non-literal word"
is a metaphor. This metaphor conveys nothing to me. If it is claimed that it
ought to convey something, one might well reply with the remark attributed to Wittgenstein "Was sich iiberhaupt erklaren lasst, lasst sich klar
erklaren." Enough on this theme.
These points might serve as contrast with some of those made earlier
about Snellgrove's work on the Hevajra-tantra. Admirably, Snellgrove
there undertook a large and risky project. He, too, makes questionable
observations118 on drang-don and nges-don (inter alia). But those observations are based on the texts, they do tell us something about their content.
The translations just discussed under (a)-(d) are advanced without argument or other evidential backing and seem to me to have simply no foundation in the texts at all. If philology is a science, such translations add
nothing to philology.
There can, of course, be no claim that my interpretations or the theoretical considerations behind them are final or immutable. On the contrary,
my interpretations are sure in due course to be replaced by better ones,
capable of accounting for the Sanskrit and Tibetan terms in a wider range
of contexts. It is equally likely that the theoretical background will be
replaced by something better. But the facts about the texts which are explained by and which become intelligible to English speakers under these
interpretations, will remain, and will have to be explained and rendered
intelligible by any improvement on them. And in this sense (only) any improvement will have to be based on what is done here. And so it is in all
scientific work.
1. R. A. F. Thurman, "Buddhist Hermeneutics," Journal of the American
Academy of Religion 46 19 (1978); Nathan Katz, "Buddhist Hermeneutics
and the Yana Controversy," to appear in the Proceedings of the 1981
Csoma de Koros Symposium (see note 3). Most of the analysis in these
otherwise very useful papers does not apply directly to the Vajrayana.
2. E. Steinkellner, "Remarks on Tantristic Hermeneutics," in the Proceedings of the 1976 Csoma de Koros Symposium (ed. Ligeti: Biblioteca Orien-
talia Hungarica no. 23, Budapest 1978). This article is based mainly on
Candrakirti's Pradipoddyotana (PPD).
3. Michael M. Broido, "bShad-thabs: Some Tibetan methods of explaining the tantras," to appear in the Proceedings of the 1981 Csoma de Koros
Symposium (ed. Steinkellner; Vienna, forthcoming). This article reviews
Tibetan Vajrayana hermeneutics on the basis of a range of sources from the
Sa-skya, Bka'-brgyud and Dge-lugs traditions.
4. Occasionally one sees the full phrase "rgyud-kyi bshad-thabs,"
"methods of explaining the tantras." But even the short phrase bshad-thabs
"methods of explanation" seems to be used only of the tantras.
5. H. P. Sastri, quoted in refs. 6,7.
6. Alex Wayman, "Concerning samdha-bhasS / samdhi-bhasa / samdhya
bhasa," in Melanges d'lndianisme a la Memoire de Louis Renou (de
Boccard, Paris 1968), pp. 789-96.
7. Alex Wayman, "Twilight Language and a Tantric Song," in The
Buddhist Tantras (Weiser, New York 1973), ch.ll.
8. Though this phrase is not offered in refs. 6-7 as a translation, it figures
importantly in the discussion.
9. F. Edgerton, Buddhist Hybrid Sanskrit Dictionary (New Haven 1953).
10. D. Snellgrove, HTT on HT Il.iii, passim.
11. V. S. Sastn, Indian Historical Quarterly (1927), p.287.
12. P. C. Bagchi, Studies in the Tantras (Calcutta 1939), p. 27.
13. Steinkellner, see ref.2.
14. See refs. 6,7; the force of "perhaps" will appear below, especially in sections 3 and 5.
15. Ref.6, 791 (twice); ref. 7, 129 (twice). See also ref. 9.
16. On the terms rjod-byed/brjod-bya and 'chad-byed/bshad-bya as used
in bshad-thabs texts, see sections 2-3 of ref. 3; cf. also note 68.
17. I owe the germ of this observation to Dr. Brian Loar.
18. See Table 1 of ref. 3.
19. See sections 3-4 of ref. 3.
20. HT II.iii.56.
21. JVS 293a2. See also refs. 6,7.
22. HT Il.iii, vv. 1, 53, 54, 55, 61, 64, 65, 66, and colophon.
23. Apart from sandhya/samdhya.
24. Refs. 6, 7, 34; HT II.iii.56 f. and HTT on it. None of these lists makes it
clear what stands for what (and note 49 of ref. 34 seems muddled on this).
25. "Madhya" in HTT is a misprint.
26. The Tib. chang means, in the first instance, "beer", but here the difference is unimportant.
27. HT II.iv.6-8.
28. (omitted).
29. E, 204a2.
30. Cf. E, 204a2, bl, b4.
31. M, 229a6, 231a3.
32. M, 229b3; the Tib. is given in note 54.
33. HTT 101; ref, 7, 134.
34. G. R. Elder, "Problems of language in Buddhist Tantra", History of
Religions 15 231-50 (1976).
35. E, 207b3; M, 234b6; the whole of L is devoted to this song.
36. HT II.vii.5-13.
37. HT II.iv.6a.
38. See K.
39. See J and M.
40. "Conventional" has two senses which can easily become confused. I am
using it just in the sense of "governed by rules" (see note 68). But it also has
the metaphorical sense of "ordinary, everyday," which is often used to
translate vyavahara (tha-snyad). The connection of sahketa with
vyavahara is well-known (cf. BHSD); but with brda it is the sense of
"governed by rules" which is uppermost, for instance in the translation of
vyakarana by "brda-sprod." Snellgrove's translation of brda by "conventional sign" at HT Il.iii.55 is confusing just because of this ambiguity of
"convention." In spite of the Skt. sahketa (instead of choma) it would be
better to translate with "secret sign" as in HT I.vii. If "conventional" means
here "governed by rules," then the qualification is redundant, for any brda
is conventional in this sense. The question is: are the conventions known to
everybody? And here, of course, they are not. Hence, "secret."
41. K, 63b5: (brda'i ngo-bo) ni/ don sgra ji-bzhin ma-yin-pa mtshon-pa'i
lus sa'am ngag-gi mam-par-gyur-ro/. This quotation of course refers to the
distinction between signs of body and of speech made in HT I.vii (most
commentaries take this either in the introduction to the chapter, or with
v.l). Apart from that, it is useful to have a definition of brda related
specifically to HT I.vii, because the Sanskrit is choma and not sahketa.
Now, sgra ji-bzhin ma-yin-pa (Skt. na-yatharuta) is a term of bshad-thabs
(see ref .3). In the bshad-thabs sections of G and H, Padma dkar-po gives
definitions of brda which, though free from the contextual constraints of
the Hevajra-tantra, are remarkably similar to the one just quoted, especially in their relation to sgra ji-bzhin ma-yin-pa. In G (Hb6) he defines the
latter and then says that it has two types, viz. brda and yi-ge. Of these,
brda is defined thus: brda zhes-bya-ba ni, dgongs-pa'i skad-nyid-kyis
gsung-pa'i 'jig-rten dang bstan-bcos phal-la ma-grags-pa'o/: "brda is something expressed precisely in sandhya-bhasa not generally known in the
world or in sastras." The context of this remark (I cannot give full details
here) helps us to see that in the remark quoted from K, don stands for
brjod-don, the communicative intention of the passage; while the contrast
between the two types of sgra ji-bzhin ma-yin (and especially Padma dkar-
po's examples) helps us to see that what is here intended is not the complete
absence of convention (which falls under yi-ge in G and H), but unusual,
non-standard conventions; of course a code falls just into this category.
The parallel passage in H (16b4) used not only the phrase ("dgongs-pa'i
skad-kyis" but also "brda'i skad-kyis." This reinforces (in respect of Padma
dkar-po) the remarks on brda-skad in section 1.5. (Padma dkar-po is
generally reliable in respect of the consistency of his terminology from one
work to another.)
42. K, 63b5-6: (brda'i nges-tshig) ni/ rnal-'byor-pa nang-gis ngoshes-shing,
gzhan-gyis mi-go-ba kla-klo'i skad Ita-bu yin-pa'i phyir, brda zhes brtags-
so/. The reference to the barbarians (kla-klo, mleccha) is found in YRM on
HT I.vii.1 and in the Vajragarbhatika (HTT 66n.l); also E, 99a6, J, 72a7.
On the YRM on choma, see also section 3.
43. HTT I.vii, passim.
44. The locus classicus is HT I.vii.l; but see E, 99ab; J, 72bl; M, 104b3.
45. (Omitted.)
46. J. 104b2: . . .tshogs-kyi 'khor-lor. . .dam-tshig mi-ldan-pa-rnams-kyi
nang-du. See refs. in n.44, also K, 63a5 and b5.
47. This theme falls under HT II.iii.65-7; J, 231b6.
48. The pithas etc. are listed at HT I.vii.10, and their connection with the
bodhisattva-bhumis mentioned at vii.ll; they are then explained in vv.
12-18. The general explanation of brda etc. is connected with I.vii.l and
II.iii.53-4, after which come the individual items of the dgongs-skad
(vv.55-60, using the classification of Kong-sprul, who calls them dgongs-
skad-rnams) and of the gsang-skad (vv.61-3). The purpose of all this is
explained in connection with either I.vii.l or II.iii.64-7. Most commentators follow this scheme.
49. M, 228b6.
50. In this passage, the many occurrences of "brda-skad" oscillate between
reference to the code in general and to the code-words as a collection of
individuals (elsewhere: brda-skad-rnams). It is sometimes rather arbitrary
whether we translate by "code" or "code-word". But later in the passage the
reference is unambiguously to the words.
51. Cf. HT II.iii.54cd. The reference is to the father-section of the anuttar-
ayogatantras. Kong-sprul here amplifies an idea which seems to lie behind
Snellgrove's translation of v.54 (HTT 99 and fn.3). But there is an important difference. According to Kong-sprul, the secret language is not a (or
the) language of smile, gaze etc.; it is the language of vows, as the tantra
itself indeed makes clear (samaya-sahketa = dam-tshig brda, 55b; dam-
tshig brda-skad, M, 229b2). This does seem to be a difference between the
Hevajra and Guhyasamaja codes.
52. Thus Kong-sprul. But HT II.iii.54cd, lit.: even by the four classes of
tantras the sandhya-bhasa is not expressed (sabdita; Tib. bsgrags-pa, made
53. M, 229b2.
54. M, 229b3: dgongs-pa'i skad-rnams rim-par gsungs-te/ de'ang brda 'di-
dag-rnams ni, phal-cher 'bras-bu'i ming rgyu-la btags-pa yin-no/
55. M, 229b6-230a6. In the same passage "brda-skad" is used of the ksetras
and upaksetras as standing for the bodhisattva-bhumis, from pramudita to
the buddha-bhumi. This use of brda-skad confirms the connection between
I.vii and Il.iii in a rather specific way.
56. But also in Bkra-shis rnam-rgyal in this context: E, 204a5 (twice); and
in Padma dkar-po (see note 41).
57. The terms dgongs-skad, gsang-skad and brda-skad are all feature-
universals (they are kinds of language). But the plural forms in -mams are
all used of items of language, i.e. words, phrases etc.
58. See the bibliography (works S1-S5).
59. Ref. 3, section 3; A, 62b-67b; E, 12a7. They also used other systems of
bshad-thabs: A, 68a-69b.
60. See refs. 2 and 3 and the sources there quoted (especially B, D, E, F, J).
61. This is not a commitment to the phrase "'intentional language" for
dgongs-skad (even as a slogan); I shall continue to use "Hevajra code" or
62. The verse is printed as given by Wayman*6'7) except that some compounds have been undone.
63. As given by Btsong-kha-pa: D, 208b3. Other authors quote slightly
different forms.
64. This word is very well-known in pramana, where it means "contrary."
One can also see directly by analysis of Btsong-kha-pa's and other texts
that it means "contrary" or "opposed" etc. and not "ambiguous."
65. The Jhana-vajra-samuccaya is an explanatory (akhya-) tantra of
Guhya-samaja, as Steinkellner correctly remarks (ref. 2, p.448-9). Elder
(ref. 34, p.236) confuses this work with D, which is Btsong-kha-pa's
commentary on it.
66. The "six alternatives" (mtha'-drug) form an important part of the
system of "seven ornaments" (rgyan-bdun). The word mtha (Skt. koti) is
here not very naturally translated by "alternative" but I have no better
suggestion. The six are usually thought of as forming three pairs of op-
nitartha (nges-don) and neyartha (drang-don)
sandhya and na-sandhya-bhasita (dgongs-bshad and -min)
yatharuta and na-yatharuta (sgra ji-bzhin and -min).
Roughly speaking, drang-don and nges-don are different interpretations of
a single passage, the nges-don one being the more advanced in the sense of
the Buddhist path to liberation; dgongs-bshad and dgongs-min are explained in the text; and sgra ji-bzhin and -min are passages understood (or
not) in accordance with normal linguistic conventions. All six, and the
attempts  to  systematize  them  in  terms  of  the  sgra/don  distinction
(Bu-ston's principle) are discussed at length in ref. 3. Some current translations of these six terms are reviewed below in section 7.
67. D, 208b3.
68. See ref. 3. The whole analytic thrust of that paper is in the direction of
showing that, in bshad-thabs, the sgra/don distinction is the one described.
But other authors have held similar views related to other contexts. For instance, in K. K. Raja, Indian Theories of Meaning, we find the remark
(p.11): "In India all the schools of thought have assumed a direct relationship between sabda and artha, which correspond to the signifiant and
signifie of de Saussure." Though the latter distinction is not clear enough
for my purposes, it is an excellent start. A related distinction is made by P.
M. Williams in translating rjod-byed(vacaka) and brjod-bya (vacya) by
"language" and "referents of language" (in "Some Aspects of Language and
Construction in the Madhymaka", Journal of Indian Philosophy 8 [1980],
1-45; see the passages quoted in notes 1 and 138-141 of that paper). The
exact distinction varies from context to context and, because of its great
importance, needs careful investigation for each context. But underlying all
distinctions of this kind is a general point which may be worth stating
explicitly. No item can function as a sign or symbol of any kind in the
complete absence of any rules as to its so functioning. For what cannot be
interpreted cannot be a sign or symbol; and interpretation presupposes
rules, without which it becomes arbitrary imputation on the part of the
interpreter. Now if meaning is held to include the rules whereby a sign is to
be interpreted, then there can be no talk of a sign being a separate item
from its meaning; for, shorn of those rules, it no longer is a sign. For this
reason, any analytical claim based on an opposition between words and
meaning, in which meaning is held to include linguistic meaning, is unintelligible. For the linguistic meaning of a word is just the rules governing its
use in the language of which it is a part; and shorn of those rules, it no
longer z's a part of that language. My account of the sgra/don distinction in
bshad-thabs,   Raja's  identification  of sabda/artha with signifiant and
signifie, and Williams' relation of rjod-byed/brjod-bya to language and its
referents, all satisfy this fundamental principle. To oppose words and
meaning is to flout the principle.
69. D, e.g. 209a5, 218a3, 209b6. The last of these is the example, where the
opposition Cgal-ba) is strengthened by the use of the verb 'geg-pa (perf.
bkag): to deny, to repudiate. See notes 74-76.
70. D, 218a4 (quoted at ref. 3, note 83).
71. D, 218a2 (quoted at ref. 3, note 77).
72. D, 209b7 (quoted at ref. 3, note 81).
73. They are desire (raga, 'dod-chags); anger (dvesa, zhe-sdang); and
bewilderment (moha, gti-mug). The use of the Sanskrit terms together in
this way probably antedates Buddhism.
74. D, ibid.
75. See note 69.
76. JVS, 292b8.
77. B, 24b5 (here taken from GST ch.5); E, 10a3 and J, 39a3 (in both cases
taken from HT II.iii.29).
78. See refs. 2, 3, 6, 7, 34.
79. B, 24b6.
80. F, 36al.
81. E, 10a2.
82. J, 39a3.
83. G, 9ab; H, 16ab.
84. But see section 6 below, and my "Abhipraya and implication in Tibetan
linguistics," to appear in the Journal of Indian Philosophy.
85. Morphologically, we do have an instrumental case-ending. What other
syntactic possibilities are there? dgongs-pas could be an adverb, after the
model of rang-gi ngo-bos grub-pa (svarupasiddha). This would give
"explained 'intentionally," which in the end amounts to much the same as
"explained by 'intention." (But it is not compatible with "in the manner of
(dgongs-pa).") On the other hand, -pas cannot be temporal; for if it were,
there would be no explanation of the fact that the basically temporal (and
not instrumental) form dgongs-nas, which is very common elsewhere,
never appears in bshad-thabs: we never find there dgongs-nas bshad-pa.
86. Padma dkar-po does this in many places (see quotations and references
in note 41). See also E, 10b2; J, 39bl.
On the vajra-song, Bkra-shis rnam-rgyal makes a subtle distinction of
level in his use of sgra ji-bzhin. The code itself is, of course, sgra ji-bzhin
ma-yin (E, 10b2). Once this unusual convention is fixed, however, its
normal use (relative to this context) is a higher-order sgra ji-bzhin: E,
208b6: /'di-dag ni, sgra ji-bzhin-pa'i yig-don-gi bshad-pa ste, drang-ba'i
don-to/. A variation on this unusual convention, interpreting the words in
terms of the sampanna-krama, constitutes a (higher-order) sgra ji-bzhin
ma-yin-pa; this interpretation is said to be due to Saroruha and others (E,
208b7: 'di-dag ni, grub-chen Mtsho-skyes-sogs-kyis rgyas-par bshad-pa'i
don'te/ sgra ji-bzhin-pa ma-yin-pa dang nges-don dang sbas-pa'i don-to/.
However Kong-sprul (M, 234b7 ff.) abandons this complication and relies
more on the tshul-bzhi (but see 236b4). A proper analysis of his treatment
will have to take into account his use of the Zab-mo nang-gi-don (for this is
the point of the words "Thams-cad mkhyen-pa Rang-'byung zhabs-kyis
gnang-ba Itar-na" at 235b5). It is not clear to me whether the Zab-mo nang-
gi-don is authoritative for Bkra-shis rnam-rgyal (it is not so for Padma
87. On the drang-don/nges-don distinction for the code in general, see E,
204b7; M, 230b4, 231b4, 233a2. Two of our authors give drang-don and
nges-don interpretations separately for the entire vajra-song: Bkra-shis
rnam-rgyal, E 207b3 and 208a7 (see note 86), and Padma dkar-po, L 2a4
and 6bl. Kong-sprul gives two similarly related interpretations, but he
classifies them as spyi'i don (M, 234b7) and sbas-don (based on Naropa, M
236a6). Our authors' hermeneutical strategies on this song deserve much
more detailed attention. But for present purposes the vital fact is that none
of them used dgongs-bshad.
88. In the La'nkavatara a typical form is sandhaya — dgongs-te, sometimes
Idem-por dgongs-te, often meaning "meaning" as a ground (not as a noun).
The form sandhaya = dgongs-te in this use is also found in the Uttaratan-
tra on 1.29 and elsewhere (see also BHSD).
89. In the Saddharma-pundarika, BHSD gives as typical forms
sandha-bhasya = Idem-por dgongs-te bshad-pa. See also the quotation in
ref. 6
90. In MSL, the important forms are abhisandhi and abhipraya (see sec. 6
for remarks on the corresponding Tibetan terms).
91. The term dgongs-pa-can, used by Bkra-shis rnam-ygal, Padma dkar-
po, Kong-sprul and others, attributes 'intention to the passage (while
dgongs-bshad is a mode of explanation in terms of 'intention).
92. The Tibetan word Idem-po (non-literal; riddle; misleading, indirect) is
very commonly used in sutra contexts and those relating to the sutras; I
have never seen it in PPD-based bshad-thabs contexts.
93. But see Steinkellner (ref. 2).
94. MSL: Levi ed., 82-3; Bagchi ed., 80-1.
95. Mdo-sde-rgyan-gyi 'grel-bshad, Sde-dge, sems-tsam: mi and tsi. The
author's name is given as Blo-brtan (short form of Blo-gros brtan-pa). The
Sanskrit title is given as Sutralamkara-vrtti-bhasyam. The vrtti on which it
comments is different from (and on the whole more detailed than) the existing Sanskrit bhasya often attributed to Vasubandhu. The section on MSL
XII. 16-18 is at mi, 240b4-243a5 and does contain a translation of most of
the existing Sanskrit bhSsya on these verses (but not of the slokas).
96. MSL, Bagchi ed. 80.28.
97. Ref. 95, mi 242b3.
98. For instance there is the passage about the interpretation of the parables
on the sands of the river Ganges and on the udumbara flower (between
Laftk. VI.6 and 7); the words a-yatharuta and upamamatra are each used
twice in the opening sentence of this passage; in other respects too it is of
great hermeneutic interest.
99. Elder's note on this gloss (40 of ref. 34) seems to miss the point of the
distinction between abhipraya and aksarartha.
100. YRM, e.g. 110.13, 117.6 (both times aksarartha-vyakhya). The word
aksarnrtja (Tib. often tshig-don) is the name of one of the tshul-bzhi (see
ref. 2, p.453). YRM seems to have had a simplified system of bshad-thabs
consisting of the following pairs: bahya-Zadhyatmikartha; nitartha/
neyartha; aksarartha/samanyartha. The first two pairs are well known; the
third is probably the first two of the tshul-bzhi. I do not know whether the
internal evidence of YRM alone is enough to enable us to work out what
Kanha himself understood by these terms. He seems not to use the other
tshul-bzhi terms (ref. 2, ibid.) garbhl and kolikam.
101. See note 83.
102. Michael Dummett, "What is a theory of meaning?": I, in S. Gutten-
plan (ed.) Mind and Language, Oxford 1975; II, in Evans & McDowell
(eds.), Truth and Meaning, Oxford 1976.
103. Donald Davidson, "Truth and Meaning", Synthese 17 304 (1967);
"Radical Interpretation", Dialectica 27 313 (1973).
104. John McDowell, (a) "On the sense and reference of a proper name",
Mind 86 159 (1977) (see also the editors' introduction to Evans and
McDowell, ref. 102); (b) "Meaning, communication and knowledge", in
van Straaten (ed.): Philosophical Subjects (Oxford 1980), p. 117 (also
Strawson's reply, p. 282).
105. See ref. 104(a), p.160.
106. W. V. O. Quine, Word and Object, M. I. T. 1960, p. 59.
107. See note 66. One might wish to restrict the context to the mtha'-drug
as interpreted by Bu-ston (B) and Btsong-kha-pa (D), since these writers are
quoted by the authors discussed here. But this restriction is pointless
because many other Tibetan writers held roughly the same view on
mtha'-drug (see E, F, J). On the other hand the restriction to bshad-thabs
(or at least to tantric) contexts must not be lifted without due care.
Especially the terms drang-don (neyartha) and nges-don (nitartha) are used
quite differently in the sutra literature (for the hermeneutics of this use, see
Thurman, ref. 1).
108. Ref. 6, 792; ref. 7, 129, 133; ref. 2, 451-6 (about 20 times!); ref. 34,
236-8 (about 10 times).
109. The description is at D, 218a3, and the illustration at D, 210a3; both
are quoted and discussed in ref. 3, section 4.3.2.
110. Ref. 6, 791-2; ref. 7, 129.
111. Ref. 2, 451; ref. 34, 236 ff. Quite apart from the unintelligibility (in the
special sense!) of these translations, they have no explanatory power. For
the drang-don explanations of the vajra-song are literal, while the nges-don
explanations are figurative (see the references in note 87; Snellgrove's use of
the underlined terms in connection with explanations of the vajra-song at
HTT 101-2 is not offered as a translation of anything and seems perfectly
reasonable; but he drifts into error on the same point at HTT 134).
112. D 218a3: gzhung gcig-la 'jug-pas.
113. see note 87.
114. Ref. 2, 451.
115. Ref. 6, 796; ref. 7, 134; ref. 2, 452; ref. 34, 236. The cautious phrasing
("When the term is coined.... it is na-yatharuta") and the restriction to
PPD in the first two of these references takes them somewhat away from
the full force of my remarks. No such caution and no such restriction are
found in the third and fourth references, however.
116. There is nothing coined about the words madana, bala. What is na-
yatharuta (sgra ji-bzhin ma-yin) is not, of course, these words, but a
certain feature of their use (viz. the convention that "madana" is to stand
for "madya", "bala" for "mamsa", etc.). And of course to say that this use
or this convention is coined is unintelligible.
117. Ref. 2, 452.
118. HTT pp.17 and 134.
Indian works
GST: Guhyasamajatantra
HT: Hevajra-tantra, ed. Snellgrove
HTT: Translation of HT by Snellgrove
JVS: Jhana-vajra-samuccaya (-tantra) — Ye-shes rdo-rje kun-las btus-pa,
Peking Bka'gyur, Rgyud Ca 290-294b
MLS:  Mahayana-sutra-alahkara  of Maitreya/Asaftga,  and its  bhasya
attributed to Vasubandhu. For the vrtti-bhasya attributed to Sthiramati,
see note 95.
PPD. Pradipoddyotana of Candrakfrti
YRM: Yogaratnamala of Kanha, printed in HT
Tibetan collections
S: Sa-skya bka'-'bum (Japanese reprint)
PK: Gsung-'bum of Padma dkar-po (Darjeeling reprint)
Tibetan works on the Hevajra-tantra
Si: Mnga'-ris-pa Jo-stan Chos-kyi Tshul-khrims (fl. c 1100-30): Brtag-
gnyis-kyi tshig-'grel (S 1 33-65)
S2: Kun-dga' Snying-po (1092-1158): Kye'i rdo-rje'i rtsa-rgyud brtag-gnyis-
kyi dka'-'grel (S 1 78-123)
S3: Bsod-nams Rtse-mo (1142-82): Kye'i rdo-rje'i brtag-gnyis-kyi mam-
par bshad-pa nyi-ma'i 'od-zer (S 2 41-109)
S4: Grags-pa Rgyal-mtshan (1147-1216): Brtag-pa gnyis-pa'i rnam-bshad
ma-dag-pa-mams 'joms-par byed-pa'i mam-'grel-dang-ldan (S 3 96-162)
S5: 'Phags-pa Blo-gros Rgyal-mtshan (1235-1280): Dpal brtag-pa gnyis-pa'i
'grel-pa-dag chung dang spyi don gsal-ba (S 6 69-83)
E: Sgam-po-pa Bkra-shis rnam-rgyal (1512-87): Dpal kye'i rdo-rje'i rgyud-
kyi rgyal-po'i 'grel-pa legs-bshad nyi-ma'i 'od-zer
K:  Padma dkar-po (1527-92):  Dpal Kye'i rdo-rje'i spyi-don grub-pa'i
yid-'phrog (PK 15 365-549)
L: Padma dkar-po: Brtag-gnyis-nasgsungs-pa rdo-rje'iglu skad-gnyis shan-
sbyar (PK 22 1-19)
J:   Kong-sprul  Blo-gros mtha'-yas (1813-99):  Rgyud-kyi rgyal-po dpal
brtag-pa gnyis-pa'i spyi-don legs-bshad gsang-ba bla-na-med-pa rdo-rje
dra-ba'i rgyan (Dpal-spungs print)
M:  Kong-sprul Blo-gros mtha'yas:  Dpal Dgyes-pa rdo-rje'i rgyud-kyi
rgyal-po brtag-pa gnyis-pa'i tshig-don mam-par 'grol-ba gzhom-med
rdo-rje'i gsang-ba 'byed-pa (Rum-btegs print)
Tibetan works on explaining the tantras (bshad-thabs)
A: Bsod-nams rtse-mo (1142-82): Rgyud-sde spyi'i rnam-bzhag (the section
on rgyud bshad-thabs-kyi man-ngag, 62b5-72b2) (S 2)
B:  Bu-ston Rin-chen-grub (1290-1364):  Dpal Gsang-ba  'dus-pa'i tikka
sgron-ma rab-tu gsal-ba (the section on de rtogs-pa'i don-du rgyan bdun
bshad-pa, 20b2 ff.) (vol. ta of the Lha-sa ed.)
D: Btsong-kha-pa (1357-1419); Dpal Gsang-ba 'dus-pa'i bshad-rgyud Ye-
shes-rdo-rje kun-las btus-pa'i rgya-cher 'grel-pa, rgyud bshad-thabs-kyi
man-ngag gsal-bar bstan-pa (Bka'-'bum vol. ta, 171b5 ff.; Otani 160
E: Sgam-po-pa Bkra-shis pnam-rgyal (1512-87): see the section on rgyan-
bdun-gyi bshad-thabs, 8b5 ff. of the work given in the Hevajra list.
F: Padma dkar-po (1527-92): Gsang-ba 'dus-pa'i rgyan zhes-bya-ba Mar-
lugs thun-mong ma-yin-pa'i bshad-pa (the section called gang-gi 'chad-
byed rgyan-bdun-gyis de ji-ltar brgyan-pa'i tshul, 32bl ff) (PK 16)
G: Padma dkar-po, Dbu-ma gzhung-lugs-gsum gsal-bar byed-pa nges-don
grub-pa'i shing-rta (the section called de spyod-pa'i gnyen-por bskor-ba
'i tshul bstan-pa, 7b3 ff.) (PK 9)
H: Padma dkar-po, Brjod-byed tshig-gi rgyud bshad-pa mkhas-pa'i kha-
rgyan (the section called tshig tshogs-pa'i rgyud-pa 'jug-pa'i tshul, 15b2
ff.) (PK 1)
J: Kong-sprul Blo-gros mtha'-yas (1813-99): see the section on rgyud spyi'i
'chad-thabs mdor-smos, 38a2 of work J in the Hevajra list.


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items