UNIVERSITY OF CAMBRIDGE ORIENTAL PUBLICATIONS 55 INDIAN SEMANTIC ANALYSIS

A series list is shown at the back of the book

INDIAN SEMANTIC ANALYSIS The nirvacana tradition

The Indian tradition of semantic elucidation known as nirvacana analysis represented a powerful hermeneutic tool in the exegesis and transmission of authoritative scripture. Nevertheless, it has all too frequently been dismissed by modern scholars as anything from folk-etymology to a primitive forerunner of historical linguistics. Eivind Kahrs argues that such views fall short of explaining both its acceptance within the sophisticated grammatical tradition of vydkarana and its effective usage in the processing of Sanskrit texts. He establishes his argument by investigating the learned Sanskrit literature of Saiva Kashmir, and explains the nirvacana tradition in the light of a model of substitution, used at least since the time of the Upanisads and later refined in the technical literatures of grammar and ritual. According to this model, a substitute (ddesa) takes the place (sthdna) of the original placeholder (sthdnin). On the basis of a searching analysis of Sanskrit texts, the author argues that this sthdna 'place' can be interpreted as 'meaning', the model thereby providing favourable circumstances for reinterpretation and change.
EIVIND KAHRS

is lecturer in Sanskrit at the University of Cambridge.

Bhairava, Nepal, seventeenth century, courtesy of the Victoria and Albert Museum, London.

I n d i a n

s e m a n t i c tradition

a n a l y s i s

T h e nirvacana

EIVIND KAHRS

C A M B R I D G E

UNIVERSITY PRESS

PUBLISHED B T E PRESS S N I A E O T E UNIVERSITY O C M RD E Y H Y DC T F H F A BIG The Pitt Building, Trumpington Street, Cambridge CB2 1RP, United Kingdom C M RD E UNIVERSITY PRESS A BIG The Edinburgh Building, Cambridge, CB2 2RU, United Kingdom http://www.cup.cam.ac.uk 40 West 20th Street, New York, NY 10011-4211, USA http://www.cup.org 10 Stamford Road, Oakleigh, Melbourne 3166, Australia © Faculty of Oriental Studies, University of Cambridge, 1998 This book is in copyright. Subject to statutory exception and to the provisions of relevant collective licensing agreements, no reproduction of any part may take place without the written permission of Cambridge University Press. First published 1998 Printed in the United Kingdom at the University Press, Cambridge Typeset in Times Norman 10/12pt in QuarkXPress® [s E ] A catalogue record for this book is available from the British Library ISBN 0 521 63188 2 hardback

In memoriam Nils Simons son

CONTENTS

Preface An outline of strategies Nirvacanasastra Praxis: Saiva Kashmir The universe of Yaska Substitution Epilogue Abbreviations and bibliography Indexes

page xiii1 13 55 98 175 268 280 295

PREFACE

From an early concern with Buddhism, in particular with Madhyamaka and Yogacara, my interest came to concentrate on the nature of the BuddhistBrahmanical controversies in the field of epistemology. However, as my research progressed, my attention shifted to the question of whether one might describe a pattern of consistencies underlying Indian Sastric exposition in general. On a larger scale, I became concerned with the questions of how classical Indian culture determined what something or someone meant or believed, how meaning was created and negotiated, how cultural change was promoted and how it was opposed. Such questions necessarily involve an extensive study of texts from various areas of Sanskrit literature, the indigenous processing of these texts, and the models and means of interpretation which were used in that processing. This led me to investigate the patterns of the Indian linguistic and ritualistic traditions. That the linguistic tradition known as vydkarana is one of the most interesting fields of study within the area of Sastric Sanskrit is common knowledge. What has been less focused upon is that there existed in India another, parallel tradition of linguistic analysis which served a different purpose. This is the tradition known as nirukta or nirvacanasdstra. Both of these traditions are classed among the six veddngas, the disciplines auxiliary to the Veda, or, more specifically, the branches of knowledge designed to preserve it. This, however, did not prevent them from being widely resorted to by Buddhists and Jains as well. While the grammatical tradition is well known to us through a number of treatises, among which the fundamental work is the Astddhydyi, Panini's famous grammar, only one basic work of nirvacanasdstra has survived, namely Yaska's Nirukta. But even though only one basic work has come down to us, the methods and principles of interpretation met with in that work remained very much alive in Indian Sastric literature. Nevertheless, in modern Indology the device of nirvacana analysis, basically a method of semantic elucidation which ultimately involves a theory of meaning (an answer to the question of what it is for words to mean what they mean), has either been disregarded completely or been interpreted as anything from folk-etymology to an ancient forerunner of historical linguistics. Such approaches leave one crucial problem unresolved: how could the method of nirvacana analysis work

xiv

Preface

for so many centuries as a highly potent tool in negotiating that intersubjective but evasive property called meaning which links words and the world? In this book I venture to show how nirvacana analysis was put to work in that ordering process whereby a culture is created and how the model underlying it fits in with patterns attested elsewhere in the Indian tradition. In broader terms, this is asking how something comes to mean what it means, rather than asking what it means. My point here is simply one of logical order: the question of how something comes to mean what it means is more fundamental logically speaking, and an answer to this question may enable us to carry our investigations of Indian cultural history a step further. That the method of nirvacana analysis did indeed become a powerful tool in negotiating meaning, I shall amply demonstrate by presenting and analysing material from the Sanskrit literature of Saiva Kashmir composed around the turn of the millennium. How it could become such a powerful tool can be explained once the method of nirvacana analysis is interpreted according to a model provided by the Indian tradition itself, namely the model of substitution: one element appears 'in the place of another. This model - known at least since the time of the Upanisads, that is to say, since the time when the Indians consciously started on a quest for meaning in the stronger sense - was refined and developed in the technical literatures of grammar and ritual. It is only through a detailed study of these literatures that it becomes possible to investigate this model of substitution and to find out what it involves. A large part of the present work consists therefore in detailed analysis of material from Sanskrit Sastric texts, an analysis which investigates the relation of 'being in the place of and makes it clear that this 'place' can be interpreted as 'meaning', a circumstance that enables ritual and linguistic elements to replace other elements under given semantic conditions. The basic ideas of this book were presented at the Colloque annuel de la Societe d'Histoire et d'Epistemologie des Sciences du Langage in Paris in January 1995. 'Interpretation et la tradition indienne du niruktd' will appear in Histoire, Epistemologie, Langage 20.1, Paris, 1998. 'Some observations on the sthdnasambandha\ which draws on material from chapter 5 of this book, will appear in a Felicitation Volume for George Cardona, edited by Madhav M. Deshpande and Peter Edwin Hook, Ann Arbor, Michigan. This book would never have been completed without support from a number of individuals. First of all I would like to express my gratitude to my ddiguru, the late Nils Simonsson, whose unfailing enthusiasm encouraged me to pursue the research that has culminated in this book. My gratitude also goes to my other teachers in Oslo, Knut Kristiansen and Georg von Simson. Albrecht Wezler was very supportive in the early stages of this work. Since we first met in 1981, George Cardona has very generously responded to any queries I have had in the field of vydkarana, within and without the scope of this book. My opponents at the public defence of my dissertation Substitution and change in the University of Oslo, Johannes Bronkhorst and Gunilla Gren-Eklund, pro-

Preface

xv

vided invaluable criticism and support. The material of that dissertation is largely incorporated here. I am also grateful to Olav Gjelsvik for stimulating discussions on philosophical matters. A Visiting Research Fellowship to Wolfson College, Oxford, in 1982-3 brought me in touch with two persons to whom I find it hard to express my gratitude in adequate terms, James W. Benson and Alexis Sanderson. Over the years, these two scholars have not only supported and criticised my work at various stages, but also generously shared with me their own ideas in their chosen fields of study. Their example has been my inspiration; the shortcomings are my own. Since 1987 this work was interrupted several times for a variety of reasons. My thanks are due to my Cambridge colleagues K.R. Norman and John D. Smith for putting me back on track, and to my wife Sudeshna Guha for making it shine. I am grateful to the President and Fellows of Queens' College, particularly to Nigel Leask, for providing a stimulating environment. Finally, I would like to thank Alison Gilderdale at Cambridge University Press for being such an excellent copyeditor when seeing the typescript through the press.

1 An outline of strategies

Traditional societies exploit flexibility while pretending permanence. This is because belief systems are not legitimated once and for all and therefore require means to cope with conflict and change without facing the challenge of admitting that these have taken place. Within traditional Indian culture, Buddhism, Jainism, and Hinduism, the last of these terms denoting what indigenous sources refer to as Vaidika, Saiva and Vaisnava forms of religion, all relied heavily on the scriptural authority of their sacred texts. To provide for change they provided scriptural change, thus advocating permanence while exploiting the exegetical potential of a text. This is why intellectual inquiry in India to a large extent developed, out of necessity, as exegesis for processing jexts. This is also why it is important to investigate the means of interpretation made use of in that processing. Exploring the Indian tradition we are to a considerable extent left to the resources of texts as our sole material. On the one hand, one may argue that this leaves us with a severe limitation of sources, in particular with regard to social status and those cultural undercurrents we have reason to believe must have existed but only occasionally surface in texts. Moreover, the only direct conclusions we are permitted to draw on the basis of texts are about ideas and idealised self-representations, whereas historical or anthropological material also touches upon those features of a culture which are not exclusively related to ideas and self-representations and are not necessarily dealt with in texts. On the other hand, the Sanskrit language served as a cultural vehicle of Indian civilisation for around three thousand years, and the same texts provide us with unique source material in that they present us with a number of conflicting belief systems and their attempts at legitimating themselves, and also with a rich theoretical literature which enables us to see what the Indians themselves thought they were doing or said they were doing. A body of individuals comprising a community builds up patterns of meaning through which it articulates its cultural identity. Such patterns of meaning are ways of classifying the world. On this view, the study of another culture becomes the study of other patterns of meaning.1 The understanding of
1

By this I do not intend to say that 'meaning' is the only gateway to the study of another culture.

a culture therefore requires translation in the widest sense, the fundamental task being the explication of cultural phenomena within the context of the patterns of meaning to which they belong. There is little in these patterns that is not specific to that culture, and meaning may exist where the outsider would not expect to find it. Although I think of meaning first and foremost as linguistic meaning, I shall nevertheless treat of meaning in any stronger sense much ift the same way, since I believe we are dealing with the same thing: a property by virtue of which language and thoughts are linked to features of the world. Our task, then, is to find interpretable patterns. But there is a serious obstacle here: there cannot be determinate meanings. Meaning is not stationary, patiently awaiting its discovery. Meaning is generated in the same way that language is generated, and thus subject to change and interpretation. In fact, meaning is created by people interpreting, be that within the patterns of meaning specific to a culture or in regard to that quest for meaning with which one sets out to study a culture.2 If there are no fixed meanings, then the interpretable pattern may change as meaning changes. Moreover, sentences of one language may be translated into another in different ways which all accord equally well with accessible empirical evidence. No one of these translations can claim to be 4the right one' for they may all be equally defensible. Famous in this respect is W.V. Quine's constructed example 'gavagai', tentatively translated 'rabbit' by an imaginary field linguist. But how are we to know that 'gavagai' applies not to rabbits, but to undetached rabbit parts or to rabbit stages? These differ not only in factual matter but also in respect of properties. This example is clearly artificial, but it serves to bring out the point.3 Quine also offers more trivial examples such as the French 'ne . . . rien' where 'rien' translates equally well as 'anything' and 'nothing' in English, and the classifiers of Japanese - and, one may add, Chinese - language.4
2

4

Perhaps most notably within the field of social anthropology, the methodological shift from function to meaning brought to light the important epistemological questions this raises; see, for example, Parkin 1982:xvi. There are other views. According to E. Gellner (1993) 'anthropologists are the anti-scripturalists of the social sciences. Obviously they are not given to the idea, tempting to at least some historians or orientalists and classicists, that there is no reality without some document or text.' In this article, which is a feature article in the TLS on the stateof-the-arts and future of anthropology, Gellner goes on to claim that 'one specific weakness in current anthropology is an excessive attention to "meaning", equivalent to the "interpretive" or hermeneutic turn in philosophy. This is the view that meaning rather than structure is the key notion of the subject.' Apart from the fact that static notions of structure are rather irrelevant, this excludes from his material the vast bulk of literary sources and the fact that human beings communicate by means of language. 'What do we need?', Gellner asks, and replies: 'We do need a language, a typology of societies and institutions' (my emphasis). Can there be a language without meaning? To focus on language and meaning is not to treat a culture as 'a selfsustaining, self-validating system of meaning'. My point is indeed the opposite: there cannot be determinate meanings, so we have to search for structures and models which take this fact into consideration. 3 For the example, see Quine 1960:51 ff.; 1969:30 ff.; 1990:42, 51. Quine 1969:33 ff.

One possible source for this indeterminacy of meaning and translation has been suggested by Quine in the form of what he calls the inscrutability of reference/ The meaning of a sentence is intimately connected with whether it is true or false. One may even argue that to know the meaning of a sentence is to know the conditions under which it is true. The sentence 'Snow is white' is true if and only if snow is white. Inscrutability of reference is the thesis that if there is one way of connecting words with objects which gives an account of truth conditions of sentences, then there will be endless ways. This theme has variants, but in its simplest form it amounts to stating that you can give exactly the same truth conditions in a language by making systematic shifts in what the singular terms refer to and what satisfies the predicates. You get a different thought if you get a different sentence to represent it. Reference, then, becomes a theoretical construct to explain how parts of sentences contribute to their truth conditions, but there is no unique relation between single external objects and words which is the relation of reference.6 This links up with a kind of holism, stressed from the beginning of Quine's work, which points to another source of indeterminacy.7 Different aspects of the mental have to fit together into large patterns, and these patterns involve normative elements. No belief can exist without being surrounded by a whole • See, in particular, Quine 1969. The theme has been developed by Donald Davidson (1979). In ? more recent work, Quine prefers to call what he is talking about 'indeterminacy of reference' (1990:50). 0 This, incidentally, deals a blow to correspondence theories of truth. Note that the view of reference outlined above is not the only prevalent view of reference in philosophical semantics. The view presented is, roughly, the position of a semantic holism advocated by Quine and later critically developed by Donald Davidson (see, in particular, Davidson 1984 and 1986). Instead of proclaiming the concept of reference as the cornerstone in a theory of semantics, Davidson suggests that a theory of semantics is an empirical theory in as much as it is a theory of truth, or, rather, that it is a theory of semantics based on truth, relying here at the outset on a Tarski-Iike concept of semantic truth for artificial languages. Roughly, A. Tarski (e.g. 1935; 1956) demonstrated what it is to call a sentence in a language L true by defining another language, a metalanguage ML, that does not have a truth-predicate but which is a truth-predicate for the first language L. According to Tarski it is not possible to extend this to natural languages because these are universal, and one will accordingly end up with the problem of self-reference. Still Davidson makes use of Tarski's concept of truth in the realm of natural languages. He renounces the claim that we must have a consistent definition, staying content with a theory of truth instead of a definition of truth. ML must still be richer than L in that it must always contain the predicate 'is true in L\ It is no longer necessarily ontologically richer, only ideally richer. Presupposing meaning, Tarski wanted to throw some light on truth. Davidson wants to throw some light on meaning, and has admirably succeeded in this although there is still a long way to go, for example with regard to the adaptation of quantificational logic, or the substitution of co-referential terms which does not retain the truth-value of sentences. Other philosophers, such as S. Kripke, K. Donnellan, H. Field, and J. Fodor retain the concept of reference as the central one in semantics. What they have in common is, roughly, an effort at giving a semantic theory empirical content by linking words and the world by a causal chain through the relation of reference. This relation between words and the world must be described without the use of intensional terms. For Davidson it is truth, for others it is reference. That it might be possible to criticise the position that the concept of reference is the central one in a theory of semantics without resorting to the almost dogmatic idea of indeterminacy or inscrutability of reference, I shall not enter into here, since the very idea of indeterminacy is instrumental in bringing forward the points I wish to make. 7 Quine 1951; cf. Davidson 1974.

4

Indian semantic analysis

galaxy of beliefs, and these beliefs fit together according to the rules, of a certain logic in as much as people try to get as consistent a picture of the world as possible. But if someone deviates from the normative pattern too much, you have reason to say that this person does not hold that belief at all. Moreover, words have semantic properties, that is to say, they do refer to things in the world. But the meaning of most words depends systematically on their relations to other words in sentences. Indeterminacy of meaning and translation may thus arise because someone means different things by the words than someone else, or means the same things but holds different beliefs. And even if the facts should happen to lead to a unique way of working matters out, you have indeterminacy in the form of inscrutability of reference. It should be pointed out that this indeterminacy of meaning and translation applies not only to translation from, say, Sanskrit into English. It applies even to meaning and 'internal translation' within the same language, even with regard to one and the same speaker. We often discover - and indeed accept, for the sake of communication - that other speakers of our language use words in ways that differ from our own. In this case we carry out 'homophonous' translation.8 Since meaning is a property which relates words and the world, then, the most striking variables in interpretation are language and ontology, the apprehension of what there is for words to mean. The inner dependency of what words mean and how the world is ordered implies that changes in one may entail changes in the other. In as much as meaning exists,.is meaning, within a social context, it is by definition something shared and, accordingly, subject to negotiation within the limits of the social context. Although meaning is subject to negotiation, then, there is nonetheless a strong element of intersubjectivity involved if meaning is to be socially accessible, which it necessarily is. This makes meanings and beliefs two sides of the same coin. There has to be a certain overlap in beliefs for people to talk about the same things, although these beliefs do not have to be identical. Still, the contents of thoughts, that is, the way beliefs, desires, and other so-called propositional attitudes are correctly described, do not depend solely on what goes on inside the head of an individual. Meaning is intimately related to usage. For example, the causal history of a word becomes part of the meaning of that word, that is, the causal relationship counts in the sense that a person's concept of a word depends on the kinds of things by which his use of words has been conditioned.
8

See Quine 1969:46. More recently Quine seems to have modified his position somewhat on this point. Having clarified his thesis of ontological relativity (see Quine 1969) by saying that what this ontological relativity is relative to is a manual of translation (Quine 1990:51), he states (ibid.:52): 'But if we choose as our manual of translation the identity transformation, thus taking the home language at face value, the relativity is resolved. Reference is then explicated in disquotational paradigms analogous to Tarski's truth paradigm . . .; thus "rabbit" denotes rabbits, whatever they are, and "Boston" designates Boston.'

An outline of strategies

5

Now, if you hold a belief, you will be surprised if you recognise that things are not as you have hitherto believed them to be. For example, if you cross the threshold between one room and another, you will be surprised if the room you enter no longer has a floor, provided it used to have one. This surprise requires some experience of a contrast between your previous belief and your subsequent belief. In other words, the surprise involves a belief about the correctness of the belief, and this belief - the concept of belief - you would not hold without language. One may argue with D. Davidson, then, that the very fact that you hold a belief requires some contrast in your mind between what you previously thought and what you find, so that holding a belief entails the concept of belief, and this contrast necessarily requires language.9 It is however crucial that we distinguish the holding of a belief from reality being different from the belief held. It seems hard to imagine a concept of belief without some concept of truth and falsity, that is, that there are both true and false beliefs. Meaning has to be intersubjective, and that is sufficient but also necessary for this contrast to arise. In some way, then, we must have an idea of intersubjective truth. The notion of truth independent of beliefs cannot exist unless there is something intersubjective, a shared standard acquired through linguistic competence. An idea of objective truth is thus an idea of intersubjective truth, and so there has to be communication if there is to be the contrast necessary for there to be beliefs. This communication is made possible by language which is thus a fundamental means of negotiation at the same time. When people share a language, they share some picture of a common world. Communication to a great extent depends on the same things being salient, but these things may be interpreted in different ways, since meaning is a negotiable entity. At the same time meaning is determined by the community in as much as there is a common language only to the extent that there is a common method of interpretation within a community. On the one hand, then, meaning is an intersubjective enterprise, while on ' the other it is always subject to negotiation and interpretation. Accordingly, a culture is not simply a set of fixed patterns of meaning, it is a web of meaning which is semantically creative and therefore rather to be described as a universe "of discourse within which there will be different interpretations of the world. It is this discourse we set out to interpret in some way or other when we set out to study a culture. Thus the object of our inquiries is nothing fixed, but rather a pattern of relationships. In order to arrive at qualified interpretations it is therefore crucial that we draw upon sources that are as wide as possible and that we look for constants within these. In order to highlight the importance and validity of the discourse metaphor, let me take one step further. If we move from the observation of, logically speaking, fairly simple things to the postulation of relatively more complex
9

See, for example, Davidson 1975. Davidson differs at this point from other philosophers of language who consider language a sufficient cause for this contrast to arise. According to Davidson it is a necessary cause.

and sophisticated things, there has to be a pattern among the simpler things. So we have to find that pattern. So-called propositional attitudes are theoretical concepts from the point of view of the interpreter. He or she can only observe the simpler things such as actions and choices, and go from there to the more complex. But no piece of logic says that if a person holds a certain belief or has a certain desire he will act in a certain way. He may even prefer not to act at all. The power of propositional attitudes to underpin and rationalise some specific way of acting may thus be merely potential until it is eventually actuated in discourse. This also implies that it is impossible to separate discourse from the notion of power, in particular the power to be semantically creative. The way we deal with this notion, however, calls for some caution. It is crucial to differentiate the notion of semantic or philosophical power from the notion of social power in as much as we are here dealing with power at two very different levels. On the semantic level we are dealing with meaning which is social and hence intersubjectively accessible. At this formal level power has nothing to do with social rules or the power to define or execute such rules. Still, if a person within a specific culture has the power to determine the reference of the words of a shared language, then that person has the capacity to interpret and determine the contents of thoughts. In one respect he or she may then have the power to define the meaning of objects and actions, even the power to define others, for this capacity rests not only on a person's ability to be semantically creative, but also on the same person's social position to be so. In this way the two notions and levels of power are linked. If there is no clearly codified system in a community, then it is possible for those who have the power to do so to reconstruct, reinterpret, and make existing data fit in as desired. In a ritual, for example, where the purpose or meaning to some extent has been lost or blurred, it is possible to encode new meaning into it; so also with scripture. A person possessing both functions of power, the social as well as the linguistic, may thus be able to orchestrate and synchronise even the syntax, the very rhythm of life in a community through such means as narrative events or ritual, or through exegesis of authoritative texts. In line with the arguments presented so far, it is clear that our object of study comes to be the universe of discourse itself, and that it is in the consistency of this discourse that it becomes possible to capture certain constants within semantically creative patterns of meaning. Now, the recognition that our matrix is the consistency of discourse does entail a basic shift in our fundamental question. Instead of asking what something means, we have to turn to a logically more fundamental question: how do we figure out what something or someone means or believes etc.? This shift from what to how is epistemological in nature, and not very different from hermeneutics conceived of as the elucidation of the conditions which make knowledge possible. It also implies that our investigation will move within the realm of cognition.

The Indian tradition and its means of interpretation Given this background it seems justified to claim that it is important to investigate the methods of interpretation employed in a specific culture and the models upon which these methods rest. The extent to which such an investigation is helpful for our understanding or, indeed, possible to carry out at all, may of course vary from culture to culture. In the case of the Indian tradition the conditions for carrying out such an investigation are remarkably favourable. Whereas the Brahmanical tradition basically is ritual and exegesis of ritual, Buddhism and Jainism present themselves as doctrines. They all represent certain conflicting ideologies and values, but they have one more thing in common: they all get their authorisation from scripture. This puts scripture in a crucial position when it comes to negotiation of meaning and authority. At the core of mainstream Brahmanism is the Veda, a corpus of scriptures considered to be revealed; at its fringes are the Agamas (lit. 'received [knowledge]'), Puranas ('old [stories and expositions]'), and the groups of heterodox texts known as Tantras. For the Mimamsaka, the extreme Vedic ritualist, even myth is arthavada, a rhetorical elaboration of topic-matter coming up in ritual, designed to encourage or to dissuade. Myth is thus part of religious discourse, its purpose being to explain and promote a rite. Jainism and Buddhism developed a canonical literature. With the rise of the Mahayana, Buddhism even produced an entirely new corpus of Sutra texts which were claimed to have been assembled by Bodhisattvas at a council of their own and therefore to be considered as genuinely presenting the Buddhist doctrine on a par with the earlier Sutra texts attributed to the historical Buddha Sakyamuni. And within the frame of the Mahayana concept of updyakausalya 'skilful means [of teaching the doctrine]' it was held that there is no right doctrine per se: the doctrine may be taught in any way that would make people understand the message, the only important things being the intention of a text and how that intention suits a particular context. All this involved exegesis and the processing of texts. When people share a belief system, then assurance of a change in belief cannot come from outside the system, nor can something inside it produce support except when it can be shown to rest on something independently trustworthy. To some extent such a trustworthy entity exists in the Indian tradition in ffrerform of established and widely accepted methods of interpretation. In fact, they have left their traces in several Asian languages far beyond the Indian borders through the translations of Sanskrit texts which accompanied the spread of Buddhism. They have permeated the Tibetan translations from Sanskrit completely, and as recently as the eighteenth century these techniques were still alive in Manchuria when the Buddhist dictionary Mahdvyutpatti was translated into Manchu. But even within such a context of fairly well-established methods of interpretation it would be a mistake to think that we study material which constitutes a logically connected whole. It is therefore important to look for

8

Indian semantic analysis

structures at a level deeper than the material taken at face value, namely to search at one remove for the models underlying these means of interpretation through a detailed examination of texts composed by those who were directly engaged in the study of scripture and performance of ritual. If we can trace such an underlying model, then we have a constant which can be investigated further in its own right in order to see how it fits in with models and patterns emerging from other parts of our material. One way to execute this task is to focus upon fixed technical methods, that is, upon systematically ordered ways of interpretation met with over and over again in the texts, and then see what patterns emerge. In this work I shall concentrate on one such consistent means of interpretation, namely the device of nirvacana analysis. By suggesting a model of thought underlying it, I hope to show how this method worked - consciously or unconsciously - in the minds of those who employed it, and how it could become such a powerful tool in negotiating and defining meaning. A note concerning method First of all, an inquiry into culturally specific means of interpretation has to be carried out through concepts and thought patterns that are clearly indigenous. This is not to advocate that form of relativism which holds that a valid explanation of another culture can only happen through indigenous concepts and thought patterns. In the attempt to achieve a basis for comparative cultural studies, I agree that such an approach runs the risk of merely replacing one set of concepts and values with another. But so does the opposite approach: an explanation of another culture in terms of concepts and thought patterns that are clearly imposed on it from the outside. For the comparative endeavour I believe one needs both and that it is indeed in the application of both sets of concepts and values that a fruitful basis for comparison lies. And if we are able to identify the mechanisms by which another culture transforms its indigenous happenings into meaningful patterns, we have achieved something that may contribute to a foundation for comparative studies. But this can only happen if we first make our inquiries into another culture through concepts and thought patterns that are recognised within the framework of that culture itself. That a better understanding of Indian culture on its own terms is required before comparative studies can be carried out with any success may be illustrated by the example of indigenous Indian grammar. Several scholars have warned against a tendency among linguists to be more concerned about how modern linguistic theories relate to indigenous Sanskrit grammar than in what that grammar actually teaches.10 Such superimposition of modern theories may
1 0

For example, P. Thieme (1961:x), R. Rocher (1968:339), C. Cardona (1969:3). The issue and the controversy arising from it have been sensibly discussed by Cardona (1976:236-7).

An outline of strategies

9

in fact block the way for a clear understanding of Indian grammar. How can any comparison be carried out before there is anything to compare? If one is to investigate systematically applied means of interpretation, it is obvious that this can only happen through an extensive study of texts. My method is therefore simply philological in that I shall - as far as possible - let the texts speak for themselves. I shall use texts as anthropologists use informants, with the additional advantage that the textual material stems from various points of a time-span covering roughly 2,500 years. If in that landscape one finds patterns that are repeated over and over again, it seems to me likely that one has detected certain basic features of classical Indian traditions, features which may then be considered stable points in our understanding and interpretation of that culture. In ordering and analysing the material I shall, of course, have to classify and arrange it according to what seems to me consistent and relevant, but I shall invariably proceed on the basis of textual material and models emerging from it. Now, if the method of nirvacana analysis is eventually to be understood only through a model revealed by the patterns that have emerged from its study, the model is inherent in the material itself. The model is thus not validated until it has emerged from an analysis of textual material which draws upon sources that are as wide as possible and then been tested on the same material from which it emerged. Although this poses some problems, among them a problem of circularity, I see no real obstacles here. For practical reasons alone only a limited number of texts can be taken into consideration. To get around the problem of having to present endless source material, I shall offer a limited range of suitable sample material from sources wide apart in time and content. In investigating further the model that emerges from the fundamental source material, I shall stick to technical literatures since this is where such models may be expected to stand out most clearly and be presented and discussed in a 'neutral' manner. Moreover, the very topic of this investigation - the study of indigenous means of interpretation and their underlying models - calls for more than a superficial study of texts. Details are what counts here. I shall therefore not only let the texts speak for themselves but in some cases let them speak at great length and provide discussion in extreme detail. This may make the book tedious to read, but I can see no other way of getting sufficiently close to the core of the textual material. Only thus will consistent patterns emerge and provide us with a model, and only thus can this model be investigated further. However, I have deliberately tried to make the book as short as possible by choosing to present - in relevant contexts - only exemplary material, that is to say, only whatever evidence I consider sufficient to make a case even where examples could be multiplied infinitely or where a case is considered inconclusive in spite of the wealth of material. The problem of circularity is solved by first exploring what tenets and models emerge from the textual material and then investigating whether - and how these eventually link up with and make sense in the light of tenets and models

10

Indian semantic analysis

known from other parts of Indian Sastric literature, the technical literatures of ritual and grammar in particular. If it is seen to fit into and link up with such patterns, this test will give the model a platform and justification outside the circle from which it emerged. At that stage it becomes possible to see how it fits the original material and my hypothesis for interpreting the device of nirvacana analysis. The role of indigenous commentaries Various positions have been taken with regard to the role that is to be assigned to indigenous commentaries in the interpretation of Sanskrit texts. With regard to the relationship between Panini, the author of the AstddhydyT, the core work of the Indian grammatical tradition, and Patanjali, the author of the Mahabhdsya or 'Great commentary' on that work, Paul Kiparsky has taken this attitude (1979:246; emphasis in the original): The real danger is that we let Patanjali ask our questions for us and thereby accept his presuppositions about the goals and theoretical foundations of linguistic analysis, instead of trying to see them in Panini's own way. Thereby we lock ourselves into Patafijali's particular universe of discourse and anachronistically impose the ideas of Patanjali on Panini, without realizing what a great gulf actually separates the two. I am aware that the case of Panini's grammar may be considered special, but to my mind it is special also because it would have been completely unintelligible without the indigenous commentaries in the first place. In other words, it is only after one has achieved an understanding of the AstddhydyT by means of the commentaries that one would be able to investigate this text at all. But I think Kiparsky's view contains tenets acceptable to many Sanskritists with regard to a number of other texts or to Sanskrit texts in general. Of course it is sometimes necessary to investigate problems in a particular text internally or to analyse a text by exploring its internal structure. Kiparsky's own work is a brilliant example of a study which poses questions with a scope beyond that of the Mahabhdsya. And of course the commentators sometimes get it wrong when they try to analyse tricky passages or consciously cover up problems when they search for an easy way out of internal inconsistencies. But such incidents should not lead one to exclude the contributions of the commentators. If we think we can ask intelligent questions, then why should we not accept that a scholar like Patanjali could? He did, after all, know the vydkarana tradition pretty well. Bhartrhari, the famous fifth-century grammarian and philosopher of language, refers in his Vdkyapadiya to Patanjali as the Master (guru)jaid the Mahabhdsya as the foundation for the justification of all principles of interpretation employed when grammatical questions are to be settled (VP 2.482): krte 'tha patahjalind gurund tirthadarsind I sarvesdm nydyabijdndm mahdbhdsye nibandhane li

An outline of strategies

II

Then, when the Mahdbhdsya, the foundation for all justifications of principles of interpretation, was composed by Patanjali, the Guru, Master of the (various] systems [of received knowledge] . . .!1 For the present investigation it is precisely the constant factors and the indigenous interpretations of them at various points in time which are of interest. The information provided by commentaries is accordingly essential. My own position in general is that commentaries should be assigned a central role in our philological investigations. For Sutra texts there is every reason to believe that oral commentaries existed alongside the texts themselves, although this is of course hard to prove. The explanations of a teacher would fulfil the same function as an oral commentary if they were transmitted regularly from teacher to pupil. And the distinction between a text that contains or embeds an afterthought - not necessarily articulated very clearly - and a text with a separate commentary seems to me quite arbitrary. Any text contains numerous levels of thought made more or less specific. The difference between some of these invariably latent but vaguely expressed levels and a commentary may accordingly turn out to be such that the commentary is closer to 'the text'. We have only to be clear about what questions we have asked on what basis and what is the textual evidence for our answers. • .-•-••"~>' "' •"*'-•.•- •-•.- ---~-> ---'-"•••One may also argue that our motivation for a study of commentaries is not in the first place linked to the interpretation of beliefs, but rather to the way in which commentaries, express cognition. A commentary is, of course, a work in its own right and rooted in its own milieu. Accordingly, it may be fairly uninteresting whether Patanjali asks questions which we consider 'intelligent'. In an investigation of the present kind it may prove more fruitful not to focus so much on the purpose of a commentator, but rather consider the methods he applied. A note on the presentation and citation of text passages Unless otherwise indicated text passages are presented in the form in which they appear in the editions from which they are cited with the obvious exception of passages from manuscripts. Emendations are indicated separately, either in notes or in the discussion of the passage in question. All text passages are cited in full and, except in a very few cases, provided with translations.
1 1

For pdtanjalind of the edition I read patanjalind; the same form pdtanjalind, although not in the context of this verse, is met with in the MSS. of Aggavamsa's Saddaniti and retained by the editor of that text, H. Smith (1928-30:710,6). However, I would also there emend to patanjalind; cf. Kahrs 1992:190, note 2. Cardona (1978:81) loo quotes the verse with the form patanjalind without further comment. I follow the interpretations of nydyabijdndm and nibandhane suggested by Cardona (ibid.:83, note 4; 87-93), partly based on the commentary of Punyaraja; tirtha 'where one goes down into water' normally refers to a religious doctrine, but seems here to be used in a broader sense.

12

Indian semantic analysis

Only thus can a reader follow the arguments presented and unambiguously judge their foundation, and only thus will my interpretations of text passages appear in an undisguised form, revealing weaknesses and preferred choices. Necessary supplements of words in the English translations which are not in the Sanskrit original are indicated by square brackets. All references to textplaces are given as indicated in the list of abbreviations and bibliography.

Nirvacanasastra

The vyakarana tradition of the Paniniyas is well known to us through a number of treatises, mainly in the form of commentaries and elaborations on the works of three grammarians, Panini, Katyayana, and Patanjali. Panini is the author or compiler of the AstddhydyT, a grammar of Sanskrit which consists of about 4,000 rules in eight chapters (adhydyas), each chapter containing four sections (pddas). To this work, pivotal to the later tradition, Katyayana added 5,000 critical remarks (vdrttikas). Finally, Patanjali commented on Katyayana and Panini in his Mahdbhdsya, the 'Great Commentary'. This monumental work is more concerned with interpretation than with the construction of a grammar which had been Panini's task a few centuries earlier, and constitutes the basis for virtually all subsequent discussions on philosophy of language and difficult points of grammar in the vyakarana tradition. Patanjali can with some certainty be placed in the second century BC but there is no conclusive evidence for the dating of either Panini or Katyayana.1 An opinion that is commonly held is that at least one century separated Patanjali from Katyayana and that at least another separated Katyayana from Panini. In stark contrast to the situation in vyakarana, only one basic work of nirvacanasastra has survived. This is the Nirukta of Yaska. Apart from this basic text, all we have is three commentaries on it and a commentary on the Nighantu, the lexicographical work upon which the Nirukta itself is a kind of commentary. Yaska Yaska is attributed with the authorship of the Nirukta; apart from that nothing is known about him. His date is uncertain, and shall probably remain so. A good summary of the rather bleak scholarly debate on this issue, which above all centres around the question of whether Yaska was prior or posterior to Panini, has been given by G. Cardona (1976:270-3). He reasonably concludes (ibid.:273): 'After all the arguments and evidence adduced in support of both
1

See Cardona 1976:260-7.

14

Indian semantic analysis

views, I think the only reasonable conclusion that can be reached at present is . . . that the question of priority remains open.' It is generally assumed that Yaska is older than Panini, but among those who have suggested arguments in favour of Yaska's posteriority we find some of the scholars most intimately acquainted with both vydkarana and nirukta? This, of course, is not intended as an argument. My own inclination is that Yaska's date falls within the later period of a possible timespan between the seventh and third centuries BCE. For most purposes the exact date is not crucial, but we would certainly get an unfocused intellectual picture if we do not pay some heed to matters of dating and relative chronology. Of some significance is the fact that Patanjali clearly attaches great importance to the Nirukta. In particular, the Nirukta plays a considerable role throughout the Paspasdhnika, the introductory chapter of the Mahdbhdsya? The commentators on the Nirukta Durga There are three extant commentaries on the Nirukta, and the problem of dating pertains to them as well.4 Durga or Durgasimha, author of a Vrtti known as the Rjvartha, was first assigned by L. Sarup (1920:50) to the thirteenth century, but on the basis of a more thorough discussion he later concluded that 'Durga can thus be appropriately assigned to the first century A.D.' (Sarup 1934: introduction, p. 101). On the evidence provided by Sarup, C. Kunhan Raja (1936:266-7) also places Durga 'long before 600 A.D.'. I have suggested elsewhere that Durga lived in the sixth century or earlier.5 My view was based on an impression of the style of Durga's philosophical language. I would certainly not object to placing him considerably earlier, but there is no conclusive evidence for doing so. Whatever the exact date of Durga may be, his work is the earliest, the most independent, and the most informative commentary on the Nirukta known to us. Skanda-Mahesvara The date of the NiruktabhdsyatTkd attributed to Skandasvamin and Mahesvara is even more problematic because its authorship raises severe difficulties too. Sarup (1934: introduction, p. 65), who worked with the only extant manuscripts of this commentary, assigned Skandasvamin to the end of the fifth For example, P. Thieme (1935; 1958:41), S.D. Laddu (1967), and M.A. Mehendale (1968). None of their arguments is conclusive, though, and several are refutable; see Cardona 1976:270-3, and S. Bhate 1968. 3 This is pointed out also by S.D. Joshi and J.A.F. Roodbergen 1986:11. 4 The dating and other information provided by H. Scharfe (1977:117-18) is extremely unreliable and provides a good example of how information which has been refuted a long time ago is still in circulation. 5 See Kahrs 1986:140-1.
2

Nirvacanasastra

15

century or the beginning of the sixth century CE, and Manesvara (ibid.:80) to the twelfth century. In a later article Sarup (1937) merely repeats this conclusion. Sarup's argument for this is as follows. Harisvamin, the commentator on the Satapathabrdhmana, mentions in this commentary a Skandasvamin, author of a commentary on the Rgveda, as his teacher:6 yah samrdt krtavdn sapta somasamsthds tatharkksrutim I vydkhydyddhydpayan1 mdm snskandasvdmy asti me guruh II My teacher is the sovereign Sri Skandasvamin who performed the seven basic Soma-sacrifices, and who taught me the Rgveda after he had composed a commentary thereon. This could well refer to our Skandasvamin. As for Sarup's dating of Skandasvamin, it rests on the interpretation of another verse found in a manuscript of Harisvarnin's commentary at Queen's Sanskrit College Library, Benares, which gives the date for the composition of Harisvarnin's work as 3740 of the Kali era. The verse runs as follows:8 yaddbddndm9 kaler jagmuh saptatrimsac chatdni vail catvdrimsat sarrids canyas tadd bhdsyam idam krtamll When three thousand seven hundred and forty years of the Kali era had passed, this commentary was composed. It should, however, be noted that Sarup (1929: introduction, p. 30) remarks on this verse as follows: The stanza giving the date of Harisvamin is not found in the edition of Samasraml. Dr. Mangal Deva, who has examined the Benares MS. describes it as 'unique' and says 'there is no reason why it (the verse giving the date) should not be regarded as genuine'. I have not seen the MS. myself and have nothing to say with regard to its genuineness or otherwise of the stanza mentioned above. I am, however, inclined to accept the stanza, recording the date, as genuine for Harisvamin can be shown to be an old commentator by independent evidence. Not only is Sarup willing to accept this as evidence, he goes quite a bit further. The verse in question informs us that Harisvarnin's commentary was written in 3740 Kali era, that is, in 638 CE. But Harisvamin also states that his ruler is a Vikramaditya, king of Avanti or Malava:I0 srimato 'vantindthasya vikramasya ksitlsituh I dharmddhyakso harisvdmi vydkhydm kurve yathdmatill I, Harisvamin, departmental head of Law for the illustrious ruler of Avanti, emperor Vikrama, am composing a commentary in accord with my own views.
6 7 9

Quoted from Sarup 1929:29. Emended from the metrically impossible vydkhyd krtvddhydpayan0 given by Sarup. I suggest that krtvd is an interlinear gloss0which can be removed. 8 Quoted from Sarup 1929:29. Emended from yadddindm. 1 Quoted from Sarup 1929:29.

16

Indian semantic analysis

The problem is that there was no such ruler of Avanti in 638. Sarup (ibid.:29) solves the problem in a simple manner: 'This gives 538 A.D. as the date of the commentary of Harisvamin, for the Kali era began on the 18th of February, 3202 B.C. The Vikrama mentioned in verse 9 [=our verse] was evidently Yasodharman of Malwa who defeated Mihiragula in 528 A.D. and assumed the title of Vikramaditya.' Now, it is generally accepted that the Kali era began in 3101 BCE, not in 3202. This would give us 638 CE and not 538. As a matter of fact, in my copy of Sarup's work (1929) the number 3202 has been meticulously corrected in black ink by hand, on a close look from 3102. This curious fact made me suspicious, so I looked up two other copies in different libraries, and, indeed, the printed numeral 1 has been carefully changed by hand to the numeral 2 in both of them as well. This is too much of a coincidence. Sarup must have realised, however, that this way out was not very satisfactory, for later (1934: introduction, p. 57) he 'solves' the problem by emending saptatrimsacchatdni of the verse in question to sattrimsacchatakdni, an emendation which is not only totally unfounded but also highly unlikely in as much as he has to insert a -ka to balance the loss of a syllable created by emending sapta- to sat-. This is the way in which he changes 3740 to 3640, thereby landing at his desired 538 CE as the date of the composition of Harisvamin's commentary. C. Kunhan Raja (1936:261) claims that 'the date of Skandasvamin can easily be decided by the mention of a date of his disciple Harisvamin', referring to the verse put forward as evidence by Sarup (1929:29). Unimpressed by Sarup's attempts at getting what he wanted, Kunhan Raja concludes (1936:262): 'Such tamperings with dates and facts do not much affect the position of Skandasvamin; and he lived about the year 600 A.D. (or 500 A.D. according to Sarup).' At least we know how he arrived at this date. No mention is made of the Vikrama king. Kunhan Raja (ibid.:266) considers Mahesvara a disciple of Skandasvamin and assigns him to the seventh century. A. Venkatasubbiah (1936), on the other hand, has argued that Sarup's date for Skandasvamin must be incorrect, and concludes that he must have lived before 1350 CE, since he is mentioned in Sayana's commentary on the Rgveda, and after ca. 1060 CE, since in his commentary on the Rgveda Skandasvamin has borrowed expositions from the commentary of Uvata who, Venkatasubbiah claims, can be dated with reasonable certainty. J. Gonda (1975:40) states that 'Indian scholars11 have attempted to determine the date of Skandasvamin, who left us a fragmentary commentary,12 at ±600 or 650 A.D.', and adds (ibid., note 12): A relative chronology can in cases such as that before us be established on the basis of quotations, references and polemics in the pertinent works: an author who is quoted by nobody else stands a good chance of being the youngest. However, these commentators do not always indicate that they are quoting from their predecessors.
1 1

Specifically, two of the three sources he mentions are Kunhan Raja 1936 and Venkatasubbiah 1936. 12 That is, on the Rgveda.

Nirvacanasdstra

17

Cardona (1976:298) states that 'Skandasvamin is probably to be placed towards the end of the fifth century', but he gives no argument for this apart from alluding to the same verse in Harisvamin's commentary (ibid., note 518) which he notes has been interpreted differently by S.L. Katre (1948) who unconvincingly places Skandasvamin in the first century CE, a date approved of by Yudhisthira MImamsaka (1965/6: introduction, p.3). Concerning the authorship of the commentary, we are also facing difficulties. Sarup (1928:9-11) lists manuscript colophons attributing certain sections to Mahesvara and others to Skandasvamin. On the basis of these colophons it may be possible, at least in theory, to arrive at some conclusions by carefully comparing the sections attributed to Skandasvamin with the ones attributed to Mahesvara. Sarup himself discarded this possibility on the grounds that the work 'bears the stamp of workmanship of a single individual' (ibid.: 12). Sarup (1934: introduction, p. 78) concluded that the commentary as we have it is the work of Mahesvara in the form of a tikd which incorporates and elaborates upon the bhdsya of Skandasvamin. Venkatasubbiah (1936:219) holds a different view, and concludes that Skandasvamin and Mahesvara were contemporaries and co-authors of the commentary. This view is shared by Kunhan Raja (1936), who, however, assigns them to a different date. The evidence for establishing the authorship is summarised by Kunhan Raja (1936:263) as follows: There are colophons in prose at the end of every section and these colophons assign the commentary sometimes to Mahesvara and sometimes to Skandasvamin (and to Sabarasvamin too). But occasionally there is a colophon in verse and this colophon assigns the work to Mahesvara. Kunhan Raja also disagrees with Sarup on the interpretation of the title Niruktabhdsyatikd, He argues (ibid.:264), I think convincingly, that the work of Yaska is called the Niruktabhdsya and that the commentary is a tiled on this work, that is to say, on the Nirukta. Accordingly, he concludes (ibid.): There is scope only for one author; it must be either Skandasvamin or Mahesvara. It may be that some portions were written by Skandasvamin and other portions by Mahesvara. Certainly the work does not contain a commentary and another commentary on this first commentary on Yaska. Clearly, on the evidence available, the dates of Skandasvamin and Mahesvara cannot be claimed as established, nor can the relationship between them when it comes to the composition of the commentary. In the present work I shall, for the sake of convenience, refer to it as the commentary of SkandaMahesvara and refer to its authorship in the plural. The most important issue for the present purpose, though, is that on the basis of all the evidence considered — if it be accepted as evidence at all Skandasvamin is later than Durga, and the commentary of Skanda-Mahesvara

J8

Indian semantic analysis

in the form we have it is later than Durga, whom it quotes repeatedly,'3 and earlier than Devarajayajvan, the commentator on the Nighantu, who is later than Sayana14 whom he quotes.l5 The vdrttikakdra Since the publication in 1931 of Mandanamisra's Sphotasiddhi with the Gopdlikd commentary of Paramesvara,16 it has been known that a metrical work entitled Niruktavdrttika existed. Paramesvara refers to it by name and quotes repeatedly from it. The thought that these quotations must stem from a commentary on the Nirukta earlier than those of Durga and Skanda-Mahesvara who sometimes attribute verses to a vdrttikakdra was first forwarded by Bhagavaddatta (1931:215). Kunhan Raja (1940-1) collated such possible quotations and suggested that since Durga quotes verses without mentioning their source and sometimes attributes them to a vdrttikakdra, and since many of them cannot be traced to any known work, they were from a hitherto unknown Niruktavdrttika. He notes (ibid.:7) that among some verses met with in Venkata Madhava's commentary on the Rgveda, three also occur in Durga's commentary on the Nirukta. However, two of these are found in the Brhaddevata; so are two verses Durga explicitly attributes to a vdrttikakdra.17 Kunhan Raja declares that the Brhaddevata is not a vdrttika, and states (ibid.:9): It is very likely that both in the Niruktavartika and in the Brhaddevata there are similar passages. In this connection it must be noticed that the passages found in Madhava are not identical with what is found in the Brhaddevata and in Durga. This may be only a recensional variation and need not lead one to conclude two separate sources. But I have to hint at various possibilities. The possibility that there might have been different recensions of the Brhaddevata and that the quotations by Durga accordingly could be from a longer recension of that text, had already been pointed out by Rajavade. Remarking on a verse attributed by Durga to a vdrttikakdra,18 he states (1921: tippani, p. 221): ayam sloko brhaddevatdydm nopalabhyate I brhaddevatdkdrdn ndnyo vdrtikakdrah, 'this verse is not found in the Brhaddevatd; there is no other vdrttikakdra than the author of the Brhaddevata'. Similarly, Rajavade (1926: tippani, p. 221) remarks: ayam sloko 'dhunopalabdhabrhaddevatdydm na vidyate, 'this verse19 is not found in the extant Brhaddevata". Introducing the Cf. Sarup 1934, appendix I, part IV, pp. 161-311): 'Parallel passages from the commentaries of Skanda-Mahesvara and Durga.' These are more than parallel passages; they are borrowings from Durga. 1 4 Sayana, the well-known commentator of Vedic texts, died in Vijayanagara in 1387 CE. 15 The problems of dating and authorship as well as the condition of the extant editions of the commentaries of Durga and Skanda-Mahesvara, I have discussed in some detail elsewhere (Kahrs 1980:33-44). 1 6 Ed. by S.K. Ramanatha Sastri, Madras University Sanskrit Series 6. 17 See below. 1 8 D 1:705,5-6; Rajavade 1921:600,2-3. 19 D 11:855,14-15; Rajavade 1926:740,12-13.
13

Nirvacanasastra

19

quotation of a verse with the words uktam ca vdrtike, 'and it has been stated in vdrttika, Durga (D 11:1050,2-4; Rajavade 1926:897,15-16) quotes a verse in the following form: madhyamd vdk striyah sarvdh pumdn sarvas ca madhvainah / gands ca same maruto ganabheddh prthak krteh. In the Brhaddevatd (BD 5.49) we find it in a slightly different form: madhyamd vdk striyah sarvdh pumdn sarvas ca madhyamahl gands ca sarve maruto gunabheddt prthak prthak. This made Rajavade remark (ibid., p. 189): durgakdle brhaddevatdaranthe bhinndh pdthd dsan, 'at the time of Durga there were different readings in the text of the Brhaddevatd'. Although he was aware of these problems, Kunhan Raja (1940-1:16) nevertheless concluded that Durga is earlier than 600 CE, and that the Niruktavdrttika must be earlier than Durga. Enthusiastically searching for manuscripts of this work, Kunhan Raja located in Kerala a manuscript of a Niruktavdrttika, or the Niruktavdrttika as he called it when he announced the discovery of the manuscript (Kunhan Raja 1943). This is the manuscript which has been edited by Vijayapala (1982) under the full title Niruktaslokavdrttika. Its author was identified by K. Kunjunni Raja (1964) as Nllakantha, to whom I shall return below. Vijayapala (1982:49) dates his work to the twelfth or thirteenth century CE. In other words, we are dealing with a work several centuries later than Durga. Arguing against Rajavade (1921; 1926) and apparently unaware of the writings of Kunhan Raja (1940-1; 1943; 1944) and his discovery, B. Bhattacharya (1950= 1958:96-109) again postulated that Durga quoted from a lost treatise entitled the Niruktavdrttika. He claims (1958:105-6):
a

Professor Rajawade states in his edition of the Nirukta ... that the Vdrttika cited by Durga is identical with the Brhaddevatd of Saunaka. The ground for this assertion is that some citations of the Vdrttika correspond exactly with the Brhaddevatd verses. That some verses attributed to the Vdrttikakdra are not to be found in the present Brhaddevatd can be explained on the hypothesis that Durga had access to a different recension of the Brhaddevatd which contained some additional verses and was apparently larger than the extant text. . . But all these arguments of Professor Rajawade cannot stand in view of the fact that verses are quoted in a commentary called Gopdlikd on the Sphotasiddhi of Mandanamisra, the great MImarisist teacher, which are all ascribed to the author of the Niruktavdrttika, none of them being traceable to the extant Brhaddevatd. That they are not traceable to the extant Brhaddevatd has a simple explanation. In 1964 K. Kunjunni Raja, who was then in possession of the Niruktavdrttika manuscript discovered by the late Kunhan Raja, made it clear that the verses quoted by Paramesvara in his Gopdlikd were identical with verses of the manuscript, whereas the vdrttikas quoted by Durga were entirely different.20
20

Kunjunni Raja 1964:251. Since the edition of Vijayapala (1982) it is possible to give exact references to the Niruktaslokavdrttika. The verses quoted in the Gopdlikd are, in order of appearance, NSV 1.6.192, 204, 205cd-6ab, 209-10, and 198cd-9ab.

20

Indian semantic analysis

Bhattacharya (1958:96-105) also attributes fourteen quotations in Durga's commentary to a Niruktavdrttika. Durga does not, but four of them he attributes to a vdrttikakdra, 'author of vdrttikas\ Bronkhorst (1985) too postulates the existence of a Niruktavdrttika known to Durga, however his argument is weak. Its foundation is the similarity between a passage met with in the Yuktidipikd (asti ndpatyam anyenajdtam sambhavatity arthah)21 and a passage from Durga's commentary on the Nirukta (ndsti seso nasty apatyam anyena jdtam).22 The context is Rgveda 7.4.7c: na seso agne anyajdtam asti. Pointing to the differences between the two statements, Bronkhorst (ibid.:91) concludes that the Yuktidipikd does not quote Durga. He goes on to claim that it remains likely that the Yuktidipikd took the sentence under consideration from a commentary on the Nirukta, since only there was this Rgvedic mantra cited in order to show that sesah means apatya 'offspring'. Bronkhorst's solution is that the Yuktidipikd quotes from the Niruktavdrttika. It is noteworthy, however, that neither of the passages in question mentions a vdrttikakdra, let alone a Niruktavdrttika. If we assume that Durga was earlier than the Yuktidipikd, which is highly probable,23 it is perfectly possible that the author of the Yuktidipikd was familiar with the commentary of Durga and based his analysis of the Rgvedic phrase on Durga's statement without quoting him verbatim. In any case, Bronkhorst's argumentation does not establish the existence of a Niruktavdrttika', it takes it for granted. This leaves us in the following situation. Nowhere does Durga or any other author mention a Niruktavdrttika which is not Nllakantha's Niruktaslokavdrttika. Durga mentions only a vdrttikakdra. Now, two of the four quotations listed by Bhattacharya (1958:97,101) which Durga explicitly attributes to a vdrttikakdra are met with also in the Brhaddevatd.2A Let me therefore consider again the suggestion made by Rajavade that at the time of Durga there might have been different recensions of the Brhaddevatd and that the quotations by Durga accordingly could be from a longer or different recension of that text. The textual history of the Brhaddevatd is complicated. The manuscripts used by A.A. Macdonell for his critical edition fall into two groups which constitute two different recensions. Macdonell
21 23 22 YD 38,9. DI 288,9 GJ Larson gives ca 600-700 as the date of the Yuktidipikd (GJ Larson and R.S Bhattacharya 1987*16), and the introduction to the summary of the Yuktidipikd in the same volume (ibid *228) states There are quotations in the Yuktidipikd from Dignaga (ca, 480-540 C E., according to Frauwallner and Hatton) and from Bhartrhari (ca., 450-510, according to Frauwallner), and it would seem that the text overall is older than Vacaspati Misra (who can be placed m the ninth or tenth century)' Mention is made of the negative evidence that the Yuktidipikd does not quote the views of Dharmaklrti concerning perception, and that it does not seem to be aware of the critique of Samkhya by Sankara The conclusion presented is that 'one is tempted to think that Yuktidipikd cannot be much later than the late seventh or early eighth century R C. Pandeya, on the other hand, cautions against accepting such negative evidence and suggests simply that the Yuktidipikd be placed somewhere between the time of Dignaga (the sixth century) and the time of Vacaspati Misra (the ninth or tenth century)' 24 (ibid) D 1.34,8-9=BD 2 102, D II 1050,3-4 = BD 5 49

Nirvacanasastra

21

(1904:xviii) points out that the longer recension contains 133 slokas not to be found in the shorter, whereas the shorter has eighteen not to be found in the longer. Macdonell's text contains 1224 slokas, of which 1073 are common to both recensions; the longer has 1206, the shorter 1091. Macdonell (ibid.) remarks: 'Continued study has, however, convinced me that the additions of the longer recension belonged, on the whole, to the original text, and that A [= the shorter recension] is an abridgement.' As for the authorship of the Brhaddevatd, Macdonell (ibid.rxxiv) concludes 'that the writer was not Saunaka himself, but a teacher of his school, who was not separated from him by any great length of time'. These circumstances do at least allow for the hypothesis that at the time of Durga there were various recensions in circulation which belonged to a tradition that goes back to Saunaka but which are not met with in the Brhaddevatd as we know it today. The question remains, however, whether the Brhaddevatd or works in a tradition that goes back to Saunaka, could fall under the label 'vdrttika\ In the classical period the term applies to two types of commentaries, either in verse or in a particular prose style. Bronkhorst (1990:142) makes the following observations: The name ' Varttika' did not only come to denote works like the Yuktidlpika and the Tattvarthavarttika. In fact, among the early works called 'Varttika' there are far more which are of a different type altogether. Most seem to follow the example of the verses quoted in the Mahabhasya, often called slokavdrttika by the commentators . . . Indeed, several works are called 'Slokavarttika'. The most famous among them was composed by the Mlmamsaka Kumarila Bhatta. Another Slokavarttika was written by Vidyananda and comments on the Tattvartha Sutra. There is also a Niruktaslokavarttika. Besides the self-styled 'Slokavarttikas' there are many 'Varttikas' which consist of verse. Bronkhorst goes on to list a series of examples, and concludes (ibid.: 143): I t is clear from this enumeration that "Varttika" came to designate primarily a commentary in verse-form.' But this is not always the case. For example, Kumarila's Slokavdrttika is not a commentary. The Brhaddevatd is also a more independent kind of work, more in the style of an Anukramanl. The fact that the Sarvdnukramani borrows heavily from the Brhaddevatd testifies to this. But the Brhaddevatd is not an entirely independent work either. Its introduction contains a classification and enumeration of the Vedic deities, and a linguistic discussion which is closely related to the Nirukta. The main body of the text is a presentation of the Vedic deities in the order of the hymns of the Rgveda. And, in the words of Macdonell (1904:xxvii), 'it contains a considerable amount of other matter, notably about forty legends, meant to explain the circumstances under which the hymns they are connected with, were composed'. On the basis of the available evidence, then, it is not possible to identify Saunaka as the vdrttikakdra with any certainty. However, there is no other likely candidate, and certainly no compelling evidence for postulating the

22

Indian semantic analysis

existence of a Niruktavdrttika as an independent, lost commentary on the Nirukta. Nilakantha As noticed already, the Niruktavdrttika manuscript discovered by Kunhan Raja has been edited by Vijayapala (1982) under the full title Niruktaslokavdrttika. Kunjunni Raja (1964:251-3) identified its author as a Nilakantha Gargya from a Yajurvedin Brahman family of Kontayur in Kerala who later became an ascetic under the name Padmapada or Padmabhagavan. Kunjunni Raja (ibid.:254-5) also makes it clear why Kunhan Raja (1944) thought the author's name was Sankara Bhagavan. Kunhan Raja had at that time not seen the manuscript in full, and the reference to Sankara Bhagavan is easily explained by the fact that Nilakantha took the ascetic name Padmapada or Padmabhagavan and as an ascetic would be referred to as Sankara; thus Sankara Bhagavan. Regarding the date of Nilakantha, Kunjunni Raja (ibid.:261-2) pointed out that all that we can definitely say is that he is earlier than Paramesvara who belonged to the close of the fourteenth century. He goes on to state that 'Nilakantha does not seem to be older than Durga or Skandasvamin, though he does not quote them', and concludes (ibid.): 'Probably Nilakantha may be one or two centuries earlier than Paramesvara.' Nllakantha's date is also considered by Vijayapala (1982: upodghdtah, pp. 45-9). Referring to Kunjunni Raja, Vijayapala places Paramesvara towards the end of the fourteenth century which is thus the upper limit for Nllakantha's date. Pointing out that the mention in the manuscript of Godavarman as the ruling king is of no avail in dating Nilakantha since it is such a common name for Kerala kings, Vijayapala instead makes an attempt to establish the lower limit by comparing passages from the commentaries of Durga and SkandaMahesvara with corresponding passages in the Niruktaslokavdrttika, He claims (ibid.:46) that Nilakantha echoes Skanda-Mahesvara, and adduces several examples to support his claim. Vijayapala concludes (ibid.:49) that there is no scent of the views presented in the Niruktaslokavdrttika in the commentary of Durga, and argues that Nilakantha refutes 'with pleasant words' views met with in the commentary of Skanda-Mahesvara without referring to them by name.25 The upper limit for Nllakantha's date is Paramesvara, and the lower limit is thus fixed by the date of Skanda-Mahesvara.26 As we have seen, this is not necessarily very helpful, and Vijayapala (ibid.) goes on to point out that there is divergence among the learned as to whether Mahesvara belongs to the twelfth century or is earlier than that, and that whether Skanda, the author of the NiruktatTkd, is identical with the author of tht Rgvedabhdsya or is more modern
25 26

durgavydkhydne tv asya paksasya gandho 'pi nopalabhyate I ato manydmahe - ndma grahanam vinaiva nilakanthah skanda-mahesvarapaksam madhurayd vdcdpratydcas tatas ca skanda-mahesvarayoh paramesvarasya cdntardle kale nilakanthah sakyate st ituml

Nirvacanasastra

23

has been a matter of dispute.27 Vijayapala (ibid.) therefore considers what he calls a middle position preferable: that Nllakantha Padmapada lived either in the twelfth or thirteenth century CE.28 He also mentions that verses belonging to the Saunaklya recension met with in Verikata Madhava's commentary on the Rgveda are quoted in the Niruktaslokavdrttika, and since linguistic usage in both is quite similar, they cannot be too far apart in time. Vijayapala (ibid.) places Venkata Madhava in the twelfth century.29 Although this argument is not convincing, it is noteworthy that Nllakantha came from a family of learned Vaidikas. I am inclined to accept Vijayapala's date as roughly correct. With Vijayapala's edition we thus possess a third commentary of the Nirukta in about 5,000 Slokas. The text is based on a single manuscript which breaks off at Nirukta 6.17. As a commentary on the Nirukta, the Niruktaslokavdrttika is a very valuable addition indeed. This lengthy exposition of the state-of-affairs of commentaries on the Nirukta, extant or lost, is intended to establish once and for all that, as far as we know, there are and have been only three commentaries on the Nirukta, in chronological order composed by Durga, Skanda-Mahesvara, and Nllakantha. To get these matters straight is important for the following reason: Durga's commentary is the earliest we have; it is also a word-for-word commentary which thus contains the whole text of the Nirukta as it existed for Durga. The Nirukta is full of later interpolations; in fact, the very nature of the text invites them. If a passage is not met with in the commentary of Durga, there is every reason to be suspicious of it. This has some bearing on the present study. More importantly, perhaps, the above discussion may serve to establish that, although scanty, there was a continuous tradition of commentaries on the Nirukta and thus a continuous tradition of nirvacanasastra which is distinct from the vydkarana tradition. The two traditions run as parallel streams. Their methods as well as their terminology remain to a large extent separate. This, of course, is not to say that they were not mutually aware of each other and that one and the same author was not familiar with or made use of both. The study of the Nirukta A more complete outline of research done on the Nirukta has been provided by me elsewhere (Kahrs 1980:24-32). Since then I became aware of the short
27

mahesvarakdlah khnstiyd dvddasasatdbdi vd tatah prdn veti vidusdm vipratipatt niruktatikdkdrah skandah rgvedabhdsyakdrdd abhinno bhinno vdrvdkkdhka ity api v padaml 28 ato 'yam madhyamah paksah sddhiydn pratiyate - khnstiydydm dvddasasatdbdydm trayodasasatdbdydm vd nilakanthah padmapddah samajdyateti I 29 api ca, venkatamddhavakrtdydm rgarthadipikdydm uddhrtdh saunakiydh slokd niruk vdrttike samdndnupurvyd samupalabhymdndh, abhiyogddisabdds cobhayatra sa 'rthavisese prayujyamdnd upodbalayanh - amu granihakdrau kdlato ndtiviprakrstdv ihl venkata-mddhavasya kdlo 'pi khnstiyadvddasasatdbdidesiyah svlkriyata aiti [sic] avidbhih I

24

Indian semantic analysis

but important contribution of Kunjunni Raja (1979). Further studies pertaining to the general nature of Yaska's work have been made by J. Bronkhorst (1981; 1984), Vijayapala (1982), Kahrs (1983; 1984), and M.A. Mehendale (1986). In order to reveal some of the background of the present study, suffice it to say that when the text of Yaska's Nirukta became accessible through the editio princeps established by Rudolph Roth in 1852 Sanskrit scholars in Europe were deeply immersed in the discoveries of comparative philology. Accordingly, they set out to study the Nirukta with the ideas of this field in mind. In fact, they also took it for granted that similar ideas were at work in the mind of Yaska. As pointed out by Bronkhorst (1981:1), August Wilhelm von Schlegel had attempted the same with Panini's grammar, which he, at least indirectly, criticised for not having taken other languages than Sanskrit into consideration. But Panini's grammar does not easily lend itself to an historical interpretation, and Schlegel's contemporaries, for example Franz Bopp and Wilhelm von Humboldt, were fully aware that the Indian Vaiyakaranas had not concerned themselves with historical problems. To look at the Nirukta through the spectacles of comparative IndoEuropean philology must have been even more inviting, Yaska's analyses in the eyes of the early European continental comparativists being primitive analyses into verbal root and suffix in the sense that every noun is derived from a verbal root. It is true that one of the fundamental features of nirvacana analysis is that all nouns are regarded as related to some activity. This activity, expressed in analysis by a finite verb or merely hinted at by the mention of a verbal root, is considered the reason for a noun signifying that which it signifies. But nirvacana analysis is indeed not simply etymological in the sense that it does not reflect the findings of historical linguistics which probably is how we understand 'etymology' today. The only way it could be considered etymological is with reference to the Greek word etymos 'true' from which the word 'etymology' itself is derived. Inevitably, such an approach imposed heavy presuppositions upon the interpretation of the Nirukta. In fact, the comparative philologists of last century made Yaska look like a predecessor in the field of historical linguistics. The Nirukta, accordingly, appeared as a kind of 'diachronic' linguistic study mainly seeking the etymology or history of a given word. Later, this interpretation was more or less adopted also by Indian scholars, foremost among them Lakshman Sarup, and it has kept its position up to the present day. The fact that it is even retained, maybe unconsciously, by a modern-day pandit such as Shivanarayana Shastri (1969; 1972), provides a good example of how cultures create each other. Ironically, it was the very discovery of the Sanskrit language which provided European philologists with the necessary material for developing the ideas of comparative philology. Indeed, one obtains a poor understanding of Indian linguistics if one

Nirvacanasastra

25

approaches that field with the ideas of Indo-European comparative philology in mind.30 By offering a new interpretation of what a nirvacana analysis is and what purpose it was meant to serve, I shall in this work amply demonstrate that the two indigenous linguistic traditions of vydkarana and nirvacanasastra are not only compatible with each other, but in fact work with similar models within a similar framework. And this basic model within which the Indian grammarians worked is really something quite different from the one at the core of Indo-European linguistics, most notably with regard to the notions of linguistic change and language development through time, as well as to the comparative idea on the whole. The background of nirvacana analysis From the Nirukta itself we know that Yaska was part of a longer tradition, although what little we know of his predecessors is based on a few scanty remarks and quotations.31 One part of the picture, however, is clear. The main intention behind a nirvacana analysis is to explain the tattva, the 'that-ness' or 'essence' of things by stating explicitly the semantic content of the words that denote them. This practice has its basis in the universe of the Vedic Indian as met with in the Brdhmana literature. The task there is to explain ritual and its implements, including formulas, acts, and material substances, by discovering their real nature and establishing their internal relations. To account for these items, their purpose and meaning in ritual, the famous equations so numerous in this later Vedic literature were made. Such an equation or bandhu 'tie; nexus' is not only established by considering the entity in itself. It is established through its name, specifically through a connection with some activity, which in these Vedic texts often is of a mythological kind. A few examples quoted by Sayana in the introduction to his commentary on the Rgveda may serve to illustrate (Say p. 29): 'tad dhutindrn dhutitvanC itil 'tarn idandram santam indra ity dcaksate* (Ai. A. 2.4.3) itil 'yad aprathayat tat prthivyai prthivitvam'32 (Tai. Bra. 1.1.3.6-7) iti ca, 'this is the ahuti-ness of the dhutis (sacrificial oblations)'; 'being truly idandra ("it-seer"), they call him Indra'; and: 'that he caused [it] to spread {aprathayat), that is the Earth-ness of the Earth'. A growing awareness of relations between linguistic elements becomes apparent already in the Vedic Samhita literature, for example in the following verse from the Rgveda (7.75.5): vdjinivati suryasya yosd citramaghd raya Tse vdsundml fsistutd jardyanti meghony usa ucchati vdhnibhir grndnall
30 32

See in this respect K Kunjunm Raja (1979), E. Kahrs (1980, 1983, 1984), and J Bronkhorst 31 (1981). See Bhattacharya 1958 62-95 for a survey of these The ed of Max Muller (1849.39) reads prthivydh prthivitvam as one would expect The TaittirTyabrdhmana, ed Mahadeva Sastri (1908 17), reads as above

26

Indian semantic analysis The young woman of the sun, carrier of the prize,33 who bestows dazzling blessings, she rules over riches, over treasures. Praised by the sages, the generous Dawn shines upon the awakening [world] besung by the performers [of sacrifice].34

The phrase usa ucchati, literally 'Dawn dawns', is a sentence in itself and establishes too much of a relation between its two elements for this combination of words to be a coincidence: the name Usas 'Dawn' is associated with the activity of shining, dawning (^vas/ucch). Fundamentally speaking, this is all there is to nirvacana analysis. Not postulating any parallels with regard to purpose and scope, I cannot help thinking of Quine's proposed reform of language with certain advantages, where a name such as 'Socrates' should be replaced by a description 'the Socratiser' where 'Socratises' is an invented predicate, and the description should then be eliminated by Russell's method, implying that a name really is an abbreviated or disguised definite description 'the x such that Fx*; Frege specifically said that such a description gave the sense of the name. The advantage is obvious, although the purposes served for formal logic were, of course, not at stake for the Vedic Indians. 'Socrates Socratises' is in any case different from a nirvacana which could certainly be called a definite description, but of a kind where both the name and the description must mean what they do because we recognise what the explanations mean and that these meanings are arrived at through the name itself and in accord with metaphysics. In the Nirukta (2.18) the above analysis appears in a technically more elaborate form: usdh kasmdtl ucchatiti satydh, 'Why [is Usas called] Usas? [Because the name is] of her who really exists so that one says "she dawns/shines" (ucchati: Avaslucch)' I shall return to the technicalities of this analysis in due course. In the later Samhita literature the attempts at creating and clarifying relations between names and activities, or between nouns and verbs, become more conspicuous. A good example is provided by the following verse from the Atharvaveda (3.13.2); consider in particular the narration of events: ydt presitd vdrunenac chibam samdvalgata I tad dpnod indro vo yatis tdsmdd apo dnu sthana II When driven forth by Varuna, you swiftly came rushing along; then Indra got hold of you as you flowed, hence you are Waters thereafter. Here the author has noticed a connection between the action (dpnot: 'obtain; get hold o f ) and the term dpah 'Waters'. He has even made an attempt at giving his explanation a logical form, and made it clear what role is played in the action by the thing signified, that is to say, as agent, object, instrument, etc. And, notably, the event is related in the past tense.35
33 35 34 On vajinivatT, see Renou EVP III 21 On vdhnibhih, see Renou EVP III 25. For further examples, also from the Brahmanas and Upanisads, see J. Gonda 1955, F Singh 1962, M A Mehendale 1978b 66-70, S Dlksa 1989, N Verma 1991, M Singh 1994, and M. Deeg 1995.

Nirvacanasastra

27

In a more explicit way such reflections are frequently met with in the Brahmanas. The Atharvaveda example reappears in the Satapathabrdhmana (6.1.1.9) in a more condensed form: yad dpnot tasmdd dpah, 'that he got hold of [them], on account of that: dpah\ This is clearly a more technical way of expressing exactly the same thought. Note that the temporality element is kept, while the event itself is presented in a more abbreviated manner. The same analysis is found in the Nirukta (9.26) in the formulation dpah dpnoteh where the nirvacana is expressed in a technical manner which I will give considerable attention further on. If the way in which I would interpret an analysis of this kind is correct (in the present case: 'dpah is of dpnotiK), then no temporality is expressed at all. In the Atharvaveda and in the Satapathabrdhmana the mythical activity which is given as the ground for using the name dpah is expressed in a past tense, thus referring to one specific event. This timelessness of the Nairuktas conflicts with the Aitihasikas, the legendarians, who refer to real events and persons in the past; the Nairuktas prefer allegorical interpretations. A quick example is found in the Nirukta, concerning the identity of Vrtra who is slain by Indra as laid out in Rgveda 1.32. Yaska remarks (Nir 2.16): tat ko vrtrahl megha iti nairuktdhl tvdstro 'sura ity aitihdsikdh,36 'but who is Vrtra? A cloud according to the Nairuktas; an Asura, son of Tvastr, according to the Aitihasikas'. The Nairuktas in general dissolve historicity. I shall later argue why I think this is so. What is really important about the Nirukta, then, is that it is the single text we possess which applies a certain method designed to give a semantic analysis of nouns, in the widest sense of that term. Moreover, the Nirukta contains lengthy discussions of linguistic and philosophical import. To Yaska a term, singular or general, is simply equipped with a meaning, and for a language user it is not necessary to keep this meaning in mind while using the term. The meaning is embedded in it, hence is objective and can be grasped intersubjectively, and nirvacana analysis provides the device for doing so. In other words, the meaning is secured through the term itself by nirvacana analysis which is considered an objective way of determining what meaning is ascribed to each word. In fact, every noun is treated as an information-invoking singular term. In Sanskrit terms the word nirvacana has been aptly defined in the following way: nirvacanam ndma sabdasya yathdrtham vyutpattih, 'what we call nirvacana is the analysis of a word in accordance with [its] meaning' (Vijayapala 1982: upodghdtah, p. 7). The Nirukta As mentioned already, nirukta or nirvacanasastra is traditionally considered one of the six veddngas (lit. 'limbs of the Veda'), the disciplines auxiliary to
36

The last phrase is omitted by several MSS

28

Indian semantic analysis

the Veda, the branches of knowledge designed to preserve it. From the Nirukta of Yaska it is clear that already by his time there were difficulties in understanding Vedic words and passages, a fact which indicates that the tradition had been broken as far as understanding the meaning of the hymns is concerned. It is to restore or achieve this meaning that the method of nirvacana analysis is outlined and put into practice in the Nirukta. Yaska describes the situation as follows (Nir 1.20): sdksdtkrtadharmana rsayo babhuvuhl te 'varebhyo 'sdksdtkrtadharmabhya upadesena mantrdn samprdduhl upadesdya gldyanto }vare bilmagrahandyemam grantham samdmndsisuh I vedam ca veddngdni ca/ The Rsis ('Seers' of the Vedic hymns) had direct access to Dharma (ritual and social duty); through [this initial] instruction they transmitted the Vedic hymns to later generations who did not have direct access to Dharma; later, inferior generations, being ill at ease with regard to the instruction, compiled this delimited work (= the Nighantu) for the comprehension [of the meaning of the Mantras] bit by bit;37 moreover [they compiled] the Veda and the Vedangas.38 In principle Yaska's Nirukta is a commentary on this compilation, a list of words known as the Nighantu (nighantuh, singular, but also frequently nighantavah, plural) or the Naighantuka. This is the earliest lexicographical work that has survived in India. Whether or not Yaska was the compiler of the entire Nighantu has been a matter of dispute,39 but I do not find this issue at all crucial in as much as there is reason to believe that the Nighantu available was not the only one of its kind. The words listed in the Nighantu are mainly Vedic, most of them occurring in the Rgveda. There are, however, a few insignificant but puzzling occurrences of non-Vedic words, for example nfrdm, listed under udakanamdni 'names for water' (Nigh 1.12), which is a word of Dravidian origin and first attested in the Mahdbhdrata. The Nighantu contains three kdndas in five adhydyas. The first, known as the Naighantukakdnda, consists of three adhydyas which give groups of synonyms where one word indicates the meaning of the whole group, as in the case of udakanamdni mentioned above. But the synonymity met with in this kdnda is not of the ordinary kind. By way of example, Nighantu 2.5 lists no less than twenty-two names for finger (angulindmdni). This would be a remarkably high number in any language, but it is even more remarkable in the present case in as much as the great majority of these words are feminine and in the plural, and the only obvious occasion for mentioning fingers would be the recitation of sacrificial formulas by the officiating priest at the sacrifice. Moreover, they show curious similarities, for example: svdsdrah 'sisters', jdmdyah 'siblings', and sdndbhayah 'having the same womb (umbilical cord)'.
37 38

I am not confident about the meaning of bilmagrahanaya, and, accordingly, about my translation. Rajavade (1940: notes, p. 290) expresses the view that vedam ca veddngdni ca is an interpola39 tion; I would not be surprised if he is right. See, for example, Skold 1926.

Nirvacanasastra

29

The second kdnda, consisting of one adhydya, is known as the Naigamakdnda or Aikapadikakdnda. It contains a list of words of more than one meaning (anekdrtha) and words of which the grammatical formation is hard to understand (anavagatasamskdra). The last kdnda, which then is also the last adhydya, is called the Daivatakdnda and gives a list of Vedic deities grouped in six sections. There is reason to believe that this list is compiled by Yaska himself.40 Outlining the purpose of nirukta (niruktaprayojanam) in the introduction to his commentary on the Rgveda, Sayana refers to the Nighantu as nirukta. He also notes that the word nighantu commonly refers to lexicographical works arranged as collections of groups of synonyms signifying a single meaning (Say p. 28): atha niruktaprayojanam ucyatel arthdvabodhe nirapeksatayd padajdtam yatroktam tan niruktaml 'gauh gmd jmd ksmd ksd ksamd1 ity drabhya 'vasavah vdjinah devapatnyo devapatnyaK ityanto yah paddndm samdmndyah samdmndtas tasmin granthe paddrthdvabodhdya pardpeksd na vidyatel etdvanti prthivindmdni etdvanti hiranyandmdni ity evam tatra tatra vispastam abhihitatvdt I tad etan niruktam trikdndaml tac cdnukramanikdbhdsye darsitam ddyam naighantukam kdndam dvitiyam naigamam41 tathdl trtiyam daivatam ceti samdmndyas tridhd sthitahll [. . .] ekdrthavdcindm parydyasabddndm samgho yatra prdyenopadisyate tatra nighantusabdah prasiddhah I tddrsesu amarasimhavaijayantThaldyudhddisu dasa nighantava iti vyavahdrdtl evam atrdpi parydyasabdasamghopadesdd ddyakdndasya naighantukatvam I tasmin kdnde trayo 'dhydydh I tesu prathame prthivyddilokadikkdlddidravyavisaydni ndmdni I dvitTye manusyatadavayavddidravyavisaydnil trtTye tadubhayadravyagatatanubahutvahrasvatvddidharmavisaydni 11 Now the purpose of nirukta is stated. That [list] wherein a class of words has been stated (-ukta) independently (nir-) [of other words] with regard to grasping [their] meaning, that is nirukta [= the Nighantu]. [That is to say,] that traditional compilation of words (samdmndyah) which has been handed down (samdmndtah),42
0

The basis for this assumption is Nirukta 7.13 where Yaska mentions certain extant lists of deities which include even their special epithets; he goes on to remark: yat tu samvijhdnabhutam sydt prddhdnyastuti tat samdmane, 'but that [appellation] which is generally recognised as a conventional [epithet of a deity] and through which [that deity] is mainly praised, that [appellation] I repeat [in my limited list]'. I ought probably to mention that the form samdmane, 1. sg. present dtmanepada, is not read by Skanda-Mahesvara who read (SM 111:72,4) samdmanet, 3. sg. optative, instead. Reading, with Macdonell (1886:60), naigamam against the editor's naigama. The words samdmndyah samdmndtah are the opening words of the Nirukta. Sarup (1921:5) translates: 'A traditional list (of words) has been handed down (to us).' The commentator Durga has a lot more to say on this phrase, and I find it worthwhile to quote the relevant passage in extenso since his wording in many respects is reflected in Sayana's words (D 1:30,1-4): gavddir devapatnyantah sabdasamuddyah samdmndya ucyatel samdnpurvasya m abhydsdrthasya karmani kdrake samdmndyah I samabhyasyate maryddaydyam iti s ndyah I sa ca rsibhir mantrdrthaparijhdndyoddharanabhutah pancddhydyisdstrasa habhdvenaikasminn dmndye granthlkrta ity arthahl

30

Indian semantic analysis beginning 'gauh gmdjmd ksma ksd ksamd" and ending 'vasavah vdjinah devapatnyo devapatnyah\43 in that text the consideration of other [words] does not take place in order to grasp the meaning of words. For all over this has been expressed very clearly in the manner of: 'so many are the names of the Earth, so many are the names of gold'. This very nirukta [=the Nighantu] has three sections. That is also shown in the commentary on the Anukramanikd:44 First the Naighantuka section, second the Naigama, and third the Daivata - thus the samdmndya is established as threefold. [. ..] The word nighantu is well known with regard to [a text] that chiefly exhibits a collection of synonymous words signifying a single meaning,45 since it is common practice to refer to such works as Amarasimha, VaijayantI, Halayudha, etc. as 'ten nighantavaK. Thus also in the present case the property of being a Naighantuka pertains to the first section since [this] enunciates a collection of synonymous words. In this section there are three chapters. As far as these are concerned, names whose scope are substances such as the Earth, and such as the worlds, the regions (space), and time, are in the first. In the second are those whose scope are substances such as human beings and their parts. In the third are those whose scope are qualities such as the property of being thin or many, or the property of being short. Sayana goes on to say that Yaska's work too is called nirukta (ibid.): pahcddhydyarupakdndatraydtmake etasmin granthe paranirapeksatayd paddrthasyoktatvdt tasya granthasya niruktatvaml tadvydkhydnam ca ' samdmndyah samdmndtaK ity drabhya ltasyds tasyds tddbhdvyam anubhavaty anubhavatV ityantaih dvddasabhir adhydyair ydsko nirmamel tad api niruktam ity ucyatel The property of being nir-ukta ('stated independently') pertains to this text [=the Nighantu], since in this text, which consists of three sections forming five chapters, the meaning (artha) of words has been stated (-ukta) independently (nir-) of other [words]. And Yaska composed an explanation of it in twelve chapters beginning

Footnote 42 (cont.) , The collection of words beginning with gauh and ending with devapatnyah is called a samdmndyah. When the kdraka karman (object) is to be denoted [the word] samdmndyah belongs to [the verbal root] mnd which has the meaning of repetition [cf. Dhp 1.976 mnd abhydse], preceded by [the preverbs] sam and dN (a). It [i.e., the samdmndya] has been repeated jointly (sam) in as much as [it presents] a limit [maryddd giving the meaning of dN; cf. A 2.1.13 an marydddbhividhyoh], thus [it is called] sam-d-mndyah. That is to say, in order that [we shall get] knowledge of the meaning of the Vedic verses, the Rsis have also made it into a text that serves as an example in the form of a compendium of the Sastra which consists of five parts in one single (ekasmin glosses the preverb sam; cf. Nir 1.3: sam ity ekibhdvam) dmndyah - 'repetition within limits'. A more pregnant interpretation of the genitives in the technical expression samdnpurvasya mndter abhydsdrthasya samdmndyah than 'belongs to' will emerge in due course later in this work. Durga accounts for the feature of 'the limit' in the following passage Where he states that there is no limit to words worth mentioning and that there would be no end to the Sastras if one were to mention all of these. The Nighantu is accordingly a limited antf unified specimen collection of Vedic words, a samdmndyah. 43 These are the first and last words of the Nighantu. 44 The passage occurs in Sadgurusisya's commentary VeddrthadTpikd on Katyayana's Sarvdnukramam on the Rgveda, ed. Macdonell 1886:60-1. 45 The expression ekdrthavdcindm suffers from the same ambiguity as the term artha itself, an issue to which I shall return below.

Nirvacanasastra

31

'samdmndyah samdmndtaK and ending 'tasyds tasyds tddbhdvyam anubhavaty anubhavatV. This too is called Nirukta. Indeed, elsewhere in his commentary it is the Nirukta of Yaska he commonly refers to as nirukta.46 The bulk of Yaska's work, then, is devoted to the analysis of Vedic words, and in keeping with his habit of analysing a word in the context of a Vedic stanza, a nigama, Yaska comments on hundreds of these. This makes the Nirukta the earliest extant commentary on Vedic texts, albeit in an unsystematic manner. The second kdnda of the Nighantu, known as the Naigamakdnda, concentrates on particularly abstruse Vedic words. In the words of Sayana (ibid.): nigamasabdo vedavdcil ydskena tatra tatra 'apt nigamo bhavatV ity evam vedavdkydndm avatdritatvdt, tasmin nigame eva prdyena vartamdndndm sabddndm caturthddhydyarupe dvitiyasmin kdnde upadistatvdt tasya kdndasya naigamatvam I The word nigama signifies Veda. For all over [the Nirukta] Vedic passages have been brought forward by Yaska in the following manner: 'and there is a nigama...'; [and] since in the second section, which forms the fourth chapter, the words that have been specified chiefly occur in that nigama [= Vedic text] only, the property of being related to a nigama (naigamatva) pertains to this section. As a commentary on the Nighantu, however, the Nirukta displays a feature which at first sight may seem curious: the first three adhydyas, the Naighantukakdnda, are hardly commented upon at all. This fact led previous Nirukta scholars to draw some rather hasty conclusions, for example Skold (1926:178): As far as I can see, these facts corroborate the conjecture which has been put forth elsewhere viz. that the first three books of the Nighantu have been amalgamated with the other parts of the same work only in a relatively late period. This conclusion is wrong, and the feature is easily explained if one keeps in mind that Yaska was primarily concerned with semantics. As mentioned above, the first three chapters of the Nighantu contain groups of synonyms, each group headed by an adhikara-word which indicates a meaning common to all the words in that particular group. From Yaska's point of view, it is therefore sufficient to determine the meaning of the adhikdra-woid.41 The two remaining kdndas, on the other hand, would require a much more detailed treatment which, indeed, they do receive. Yaska states (Nir 1.15) that 'without it (= nirvacanasastra) a clear understanding of meaning in respect of Vedic stanzas is not possible; nor is there a proper indication of accent and grammatical formation for someone who does
46

47

E.g. at RV 1.158.2: tathd ca niruktam - 'gaur itiprthivyd ndmadheyam yad duram gatd bha yac easydm bhutdni gacchanti gater vaukdro ndmakaranah' iti. This is clearly a quotatio from Nirukta 2.5. Examples could be multiplied. For a more lengthy discussion of this, see Kahrs 1984:149-50.

32

Indian semantic analysis

not have a clear understanding of the meaning' (idam antarena mantresv arthapratyayo na vidyatel artham apratiyato ndtyantam svarasamskdroddesah). This makes him go on to characterise his branch of knowledge (vidydsthdna) as vydkaranasya kdrtsnyam svdrthasddhakam ca, 'a complement to grammar, moreover something which is a means to its own end' (ibid.), implying thereby that nirukta takes care of that part of linguistic analysis which vydkarana does not cover. Alternatively, vydkaranasya kdrtsnyam could be taken quite literally as 'the totality of grammar', thus also in this case implying that nirukta serves to make grammar complete. Claiming that to Yaska grammar would to a great extent coincide with Vedic grammar, S.D. Joshi and J.A.F. Roodbergen (1986:12) even suggest that the phrase could be interpreted 'to mean, not that the Nirukta is the whole of grammar, but, rather, that it serves to make Vedic grammar complete'. Although I do not think this is directly wrong, I do not think that Yaska's statement necessarily is intended to be that way delimited. Illuminating in this respect is the following statement (Nir 2.2): athdpi bhdsikebhyo dhdtubhyo naigamdh krto bhdsyantel damundhl ksetrasddhd itil athdpi naigamebhyo bhdsikdhl usnaml ghrtam iti, 'moreover, Vedic primary nouns are analysed on the basis of verbal roots belonging to classical Sanskrit, such as damilndh, ksetrasddhdh, but also classical [primary nouns] on the basis of Vedic [roots], such as mnam, ghrtam\ The intricacies of these roots and the meaning of the words adduced as examples need not concern us here. What is interesting about this statement is that it indicates that for Yaska bhdsd and chandas or naigama were two domains of the same language, but not linguistically separate in time. In other words, he would not treat of them differently. Yaska, who was primarily concerned with the analysis of words which by vydkarana would be considered avyutpanna 'underivable', simply set out to analyse 'also Vedic words, the grammatical formation of which is not immediately clear' (anavagatasamskdrdms ca nigamdn; Nir 4.1). Accordingly, he was not very interested in words which yield their semantic content through a regular morphological analysis, though he is in no way opposed to grammar. As a matter of fact, the two methods overlap. Conversely, grammarians make frequent use of nirvacana analysis. For example, the famous grammarian Patafijali makes use of it in his Mahdbhdsya, and the method is a design feature in the twelfth-century Pali grammar Saddaniti of the Burmese monk Aggavamsa. With regard to the relationship between vydkarana and nirukta, the Tibetans speak of two traditions of linguistic analysis, sgra bzin du 'according to sound' and don bzin du * according to meaning'. This is reflected also in a statement quoted to me in December 1993 in Madras by Agnihotram Ramanuja Tatachariar, one of the living authorities of Vaidika Srauta religion: sabdapradhdnam vydkaranam, arthapradhdnam niruktam, 'vydkarana has linguistic items as its predominant concern, nirukta has meaning as its'. Sayana, who generally maintains a tradition of ritualistic Veda interpretation, faces the problem of a broken tradition as far as the meaning of the Veda

Nirvacanasastra

33

is concerned. He is quite clear about the relationship between vyakarana and nirukta, and frequently quotes Yaska's nirvacanas although he may suggest alternatives of his own. And sometimes Sayana has based his exegesis on the Nirukta, although he fails to make this clear. The following example will suffice to demonstrate this. Sayana glosses aksanvdntah of RV 10.71.7 by aksivantah 'who possess eyes', the word aksi 'eye' being explained as follows: anena drsyate sarvam ity aksi I yad vdl taijasatvdt anyebhyo 'rigebhyo vyaktataram, 'everything is seen by means of it, hence [it is called] aksi\ or else: because it is made of fire (tejas) it is more visible (vyaktataram) than any other part of the body'. As nirvacanas these explanations do not make immediate sense. They do, however, when their reliance on the Nirukta is made clear. Sayana first interprets the word aksi as karanasddhana, that is to say, as the instrument of the act of seeing, the verbal root ^drs 'see' being a synonym of ^Icaks by means of which the word is analysed in the Nirukta (Nir 1.9). There Yaska also cites an alternative analysis, attributed to Agrayana, by the verb anaktih (^afij). Among the meanings attributed to the verbal root yanj in the Dhdtupdtha (Dhp 7.21 ahju vyaktimraksanagatisu) is vyakti 'make appear; reveal', hence Sayana's vyaktataram. In this way the Nirukta has infused a whole tradition of Vedic exegesis. And Sayana is quite unambiguous when it comes to the purpose served by nirvacanasastra (Say p. 29): evam granthakdrenoktdh tattatpadamrvacanavisesdh tattanmantravydkhydndvasare evdsmdbhir uddhansyante I na ca nirvacandndm nirmulatvam sankanlyaml etadvyutpattyartham eva brdhmanesu padamrvacandndm kesdmcid uktatvdtl [...] granthakdro 'pi tatra tatra svoktamrvacanamulabhutabrahmandny uddhansyati I kesdmcin nirvacandndm vydkaranabalena siddhdv api na sarvesdm siddhir asti I ata eva granthakdra aha - 'tad idam vidydsthdnam vydkaranasya kdrtsnyam svdrthasddhakam ca' itil tasmdt veddrthdvabodhdyopayuktam niruktaml Thus various nirvacanas of this or that word which have been stated by the author [= Yaska] will be quoted by me on the occasion of interpreting this or that mantra. And there is no reason to suspect that nirvacanas are unfounded. This is because some nirvacanas of words have been stated in the Brahmanas for the sheer purpose (artha) of derivational analysis (vyutpatti) of them. [. ..] Also the author [=Yaska] will all over be found to quote Brahmanas in support of nirvacanas he himself has stated. Even though some nirvacanas are established on the strength of grammar (vyakarana), not all are established [by that means alone]. And for this very reason the author [=Yaska] has stated: 'this branch of knowledge [= nirukta] is a complement to grammar, moreover something which is a means to its own end'. Therefore nirukta is useful for grasping the meaning (artha) of the Veda. Sayana, then, uses the authority of Veda (Brahmanas) to argue for the validity of nirvacana analysis as an exegetical tool. In fact, Sayana says, although some nirvacanas are established on the strength of vyakarana alone, nirvacanasastra is needed as a supplement. And, significantly, to Sayana the purpose served by nirvacanasastra is to establish the meaning of Vedic

34

Indian semantic analysis

Mantras. But, as will be evident further on, the Brahmanical tradition did in no way delimit the device of nirvacana analysis to the meaning of Vedic texts alone. Indeed, as an exegetical tool nirvacana analysis is not even confined to the Indian tradition and Indian languages. The method permeates the translations of Sanskrit texts into other Asian languages through the spread of Buddhism. Suggestive examples from the Tibetan translations have been offered by N. Simonsson (1957:238-80). I shall confine myself here to a single example from the Chinese. In Paramartha's 48 translation of Vasubandhu's Abhidharmakosabhdsya we meet with a striking translation of the word sdstra 'teaching; doctrine', in this context the Buddhist teaching and more specifically the Abhidharmakosa itself.49 The common derivation of sdstra is from the root sds 'instruct; teach' by the instrumental suffix -tra, in the sense of 'something by means of which one teaches'. This is no longer Paramartha's basis for translation. Here - and remember that Paramartha was an Indian who no doubt knew his vydkarana and nirukta well - the word is rendered by the two characters W* mieand dk ji. The first means 'extinguish; exterminate; destroy', the second 'to cross a stream; aid; relieve'. The word mie is commonly used in the translation of nivrtti or nirodha, the cessation of suffering in the context of the third so-called noble truth of Buddhism; also of nirvana, as blown out, extinguished; but above all of nirodha in the sense of the destruction or suppression of the passions, or klesas - the defilements of the mind by virtue of which actions (karman) gather force and fruitfulness and necessarily produce retribution, notably a new existence. An aim in Buddhism is accordingly to destroy these passions and melt away all traces of them (vdsand).50 The word ji has the double meaning 'cross over' and 'aid; save' - to cross over the sea of existence to nirvana is a common image in Buddhist scriptures. Now, how does this come about? Paramartha has split the word sdstra into its two syllables sds and tra. The first he relates not to Asds 'instruct' but to Asas 'destroy'. The second part, tra, he relates either to Atrai (trdyate, trdti) 'save; rescue', or to At? 'cross over' (tarati) and its causative tdrayati 'cause to
48

49

50

Paramartha was an Indian Buddhist monk who arrived m China in the year 546 AD He stayed for two years m Guangzhou (Canton), then went on to the Liang capital where he was well received by the emperor Wu The emperor wanted to establish a translators' bureau under Paramartha's leadership, but political turmoil put an end to this plan, and Paramartha spent the rest of his life travelling around in China, homesick and miserable to the point that he even attempted suicide Saved by his disciples, he died of illness instead a year later, in 569 But during his years in China he managed to translate a large number of Buddhist texts into Chinese, among them the Abhidharmakosabhdsya of Vasubandhu Taisho 1559, vol 29, p 161c That we are dealing with a translation of the word sdstra and not of (sisya-)sdsana as maintained m the index of A Hirakawa et al (1973 351 sdsana-, 352 sisya-) will become clear from the following exposition Indeed, this is a good example of how important it is to master nirvacanasdstra for any kind of lexicographical work Cf,eg,E Lamottel974

Nirvacanasastra

35

cross over; save; rescue'. The root Vff (Dhp 1.1018 tf plavanataranayoh) would neatly carry the same double meaning as ji, but the problem with this root is that it would not give us tra without some morphological violence. Still, the parallel double meaning must be more than a coincidence. In any case, we have here a full interpretation of the sdstra, the Buddhist doctrine as expounded in the Abhidharmakosa, as that which destroys [the klesas] (basically, impedes ignorance) and therefore saves, that is, makes one cross over the sea of existence. No wonder Chinese Buddhist texts can seem hard to Sinologists. Indeed, any reader of Sanskrit texts who is not familiar with the devices of nirvacanasastra will soon face textual difficulties. Yaska's exposition of nirvacanasastra: a brief outline As mentioned already, one of the fundamental features of nirvacana analysis is that all nouns are regarded as related to an activity expressed in language by a verbal form. Yaska formulates it thus (Nir 1.12): tatra ndmdny dkhydtajdmti sdkatdyano nairuktasamayas cal na sarvdniti gdrgyo vaiydkarandndm caike, 'with regard to this [=the four classes of words], nouns arise because of [the actions denoted by] verbs, according to Sakatayana; this is also the doctrine of the Nairuktas; not all [nouns], according to Gargya and some of the Vaiyakaranas'. This is how I think this statement is to be interpreted, and certainly not as giving support to the 'historical interpretation' of the Nirukta. The arguments for this will be developed throughout this book. Suffice it here to say that Yaska himself states that his purpose is a semantic one. That this purpose is achieved through an analysis into linguistic elements which, when possible, agree with the grammatical derivation of a word, is hardly to be held against it. With this in mind, let me now turn to some of the general principles for nirvacana analysis as laid down by Yaska in the second chapter of the Nirukta. The opening words of his exposition are (Nir 2.1): atha nirvacanam, 'now [follows an exposition of] nirvacana [analysis]'. The word nirvacanam here is defined by Durga in the following manner (D 1:154,18-19): apihitasydrthasya paroksavrttdv atiparoksavrttau vd sabde nihkrsya vigrhya vacanam nirvacanam, 'nir-vacanam is a statement {vacanam) of the hidden meaning, extracting {nihkrsya = nir-) [it], that is, isolating [it], m a word of which the grammatical formation is out of sight (unintelligible) or completely out of sight (completely unintelligible)'. Such a statenienlXy^ggngm) consists lnjtnjanj^s^ ^ is called thereby e s t a b l i s h i n g \ 3 J ^ ^ ^ ^ ^ ^ x ^ ^ h ^ j ^ i ^ ^ ^ may be a simple fasETasin the case of a word which lends itself to a straightforward grammatical derivation; it may also be quite complicated. The terms paroksavrtti and atiparoksavrtti, referring to words with grammatical formations that are hard to figure out and extremely hard to figure out, respectively, are commonly

36

Indian semantic analysis

resorted to by the commentators. In addition to these they operate with a third category: words which are pratyaksavrtti 'of an evident formation'. It is with reference to this latter category that Yaska continues his exposition (Nir 2.1): tad yesu padesu svarasamskdrau samarthau prddesikena gunendnvitau51 sydtdm tathd tdni nirbruydt, 'so, in the case of words where accent and grammatical formation would be in agreement with the meaning {samarthau) [which is to be expressed] [and are] accompanied by a [phonetic] quality which is in accordance with the grammatical derivation, they should be analysed in a regular manner'.52 In other words, this lays down that in principle one should go for the obvious first. If a word seems to contain, an element which in an obvious way links it to a verbal root and another element which is known as or looks like a suffix, this is how the word should be analysed and its concealed meaning revealed. As pointed out by Mehendale (1978b:73), one may also include all the words dealt with by the grammarians in the Unadisutras under the scope of this rule. Matters are not always that straightforward, so Yaska goes on to say (ibid.): athdnanvite 'rthe 'prddesike vikdre 'rthanityah parlkseta53 kenacid vrttisdmdnyena, 'but when the meaning is not accompanied [by a regular accent and grammatical formation] [and a phonetic] modification is not in accordance with the grammatical derivation, one who is intent on a meaning should examine [the word] through some similarity with a [phonetic] formation [accepted by the grammarians in other cases]'. This means that in cases where the accent and grammatical formation are not regular and therefore do not reveal the concealed meaning of a word, and the phonetic modifications of the stem do not give any indication of a verbal root, one should proceed to analyse, with a particular meaning in mind, on the basis of similarity with a phonetic change that has been accepted by the grammarians for the explanation of some
51

Samp (1927) reads vikarendnvitau, but the v.l. gunena0 occurs frequently and is also the reading of Durga (D 1:155,3) and Nllakantha (NSV* 1.4.18). It is discussed as a variant (and indeed read by one Ms) in the commentary of Skanda-Mahesvara (SM 11:2,4-5). If Sarup's reading is preferred one would have to translate '... accompanied by a [phonetic] modification which is in accordance with the grammatical derivation'. On this evidence, there is every reason to emend Sarup's text. What might speak against doing so is the continuation of Yaska's text: athdnanvite 'rthe 'prddesike vikdre 'rthanityah parikseta kenacid vrttisdmdnyena below). Were the reading vikarendnvitau retained, it would provide a rather neat parallelism with this subsequent passage. 52 The cumbersome wording and my translation here is due partly to the previous statement at Nirukta 111 {tad yatra svarasamskdrau samarthau pradesikena vikarendnvitau [or, p ably, v.l. gunendnvitau] sydtdm), partly to the cryptic formulation and the words chosen; the interpretations and translations among commentators and modern scholars vary; see Mehendale 1978a (= 1978:57-62) and 1978b:73, and the modifications of Mehendale's interpretation suggested by Bronkhorst (1984:5-6). The translation of prddesikena as 'in accordance with the grammatical derivation', or 'which fits the [grammatical] derivation' (Bronkhorst) is based on an interpretation of pradesa and hence of prddesika suggested by Scharfe (1977:121-2). If this is not accepted, I suggest one interprets prddesikena gunendnvitau as *[and] accompanied by a [change in the phonetic] quality [of the stem caused by a grammatical operation] which is indicating [the activity expressed by a verbal root]'. 53 Sarup (1927) inserts a danda here which is totally unwarranted.

Nirvacanasastra

37

other form elsewhere. In other words, this rule reflects the situation in which one is facing a formation referred to by the commentators as paroksavrtti 'out of sight (unintelligible)'. It is difficult to figure out, but by analogy with similar phenomena elsewhere in grammar it might be done.54 The situation may be even more difficult, so Yaska continues by laying down a principle for how one should deal with a form referred to by the commentators as atiparoksavrtti 'completely out of sight (completely unintelligible)' (ibid.): avidyamdne sdmdnye 'py aksaravarnasdmdnydn nirbruydt, 'even when [such] a similarity [with a phonetic change accepted by the grammarians in other cases] is not found, one should analyse on the basis of [a possible similarity] in syllables or in single sounds'.55 He underlines this by adding (ibid.): na tv eva na nirbruydt, 'but never indeed should one not analyse'. These three basic rules, crowned by the statement that one should never give up analysing, lay down the most fundamental principles of nirvacanasastra. They also display an increasing slackness with regard to a strict application of the rules of grammar, a fact Yaska makes unambiguously clear when he goes on to remark (ibid.): na samskdram ddriyeta I visayavatyo hi vrttayo bhavanti, 'one should not pay attention to the grammatical formation, for [phonetic] changes possess a wide range56 [of possibilities]'. That is to say, one should try to stick to the rules of grammar as far as possible, but if this is of no avail in bringing out the hidden semantic content of a word one should abandon them
54

This rule too has been subject to a great variety of interpretations and translations. My interpretation of vrttisdmdnya above is the one suggested by Mehendale (1978a) who has discussed the passage and the various attempts at interpreting it in great detail. Mehendale argues as follows (ibid.:61—2): 'The word vrttisdmdnya seems to mean "commonness of behaviour (of a given sound), i.e. the phonetic change undergone by it". What Yaska means to say is that m cases of difficult words - i.e. words, the accent and the grammatical formation of which are not regular (ananvite arthe) or in cases of words where the word to be derived is not indicative of any action (aprddesike vikdre), e.g. a word like indra - in such cases one should first take a root similar in (sound and) meaning with the word to be derived and derive it from that root on the basis of the commonness of behaviour in undergoing a phonetic change, that is he should try to see if the phonetic change implied in such a derivation has any similarity with a phonetic change accepted by the grammarians for the explanation of some other forms in the grammar.' He adduces the examples of indra and hasta. If indra is related to the root indh, one would have to account for the change of dh > d\ if this can be observed in acceptable grammatical formations, for example dadhdti and not *dhadhdti in the reduplicated present form of ^Idhd, there is some justification in accepting indra for *indhra as related to ^indh. Similarly, if hasta is to be related to han, a phonetic change n > s has to be accounted for by finding some similar change elsewhere in grammar. Mehendale concludes (ibid.): 'This is what is really meant by vrttisdmdnya, viz, commonness of behaviour (of sounds).' Bronkhorst (1984:4) reviewing Mehendale (1978a) argues that 'it seems better to translate ananvite 'rthe "when the meaning is not accompanied [by the right accent and formation]" and apradesike vikdre "when the modification is not such as fits the [grammatical] derivation"'. What is lost when the passage is taken in this way (which is what I have done above) is the parallelism between prddesikena gunendnvitau (or even vikdrendnvitau) of the preceding rule and the wording ananvite 'rthe 'pradesike vikdre in the present one. 55 Again I follow the interpretation of Mehendale (1978a), assuming sdmdnye = vrttisdmdnye; see preceding note. 56 Durga (D 1:157,6) glosses visayavatyah by bahusamsayavatyah 'possessing lots of doubt'.

38

Indian semantic analysis

and proceed by analysing on the basis of similarities in syllables and single sounds. Accordingly, Yaska concludes the passage by adding (ibid): yathdrtham vibhaktih sannamayet, 'the divisions [into syllables and single sounds] should be interpreted according to meaning'. This statement does not necessarily imply that he considers syllables and single sounds to,be meaningful units, a view that is certainly not shared by vydkarana; maybe all it says is that someone who is intent on a particular meaning (arthanityah) should keep this in mind when setting out to analyse a word on the basis of syllables and single sounds. On the basis of his general practice, however, I am inclined to conclude that Yaska considers syllables and single sounds to be meaningful units. Another important statement is the following (Nir 2.7): tdni cet samdnakarmdni samdnanirvacandni I ndndkarmdni cen ndndnirvacandni I yathdrtham nirvaktavydni I itimdny ekavimsatih prthivindmadheydny anukrdntdni, 'if the activities [associated with a word] are uniform, the analyses are uniform; if the activities are multiform, the analyses are multiform: [and] they should be analysed according to meaning; with this statement in mind, the twenty-one words for Earth are dealt with'. That Yaska is all in favour of grammar and that he deems it essential that a nairukta is well versed in vydkarana is evident from the following statement (Nir 2.3): naikapaddni nirbruydtl ndvaiydkarandyal ndnupasanndyal anidamvide vdl nityam hy avijhdtur vijndne 'suydl upasanndydtu nirbruydtl yo vdlam vijndtum sydtl medhdvinel tapasvine vd, 'one should not analyse single words57 to someone who is not trained in vydkarana, who has not for57

The exact meaning of the term ekapada is unclear. Mehendale (1978b:75) reads naikapaddni nirbruydt as a separate statement and interprets: 'one should not attempt etymologies of single words without knowing their context'. The technical term ekapada occurs only twice in the Nirukta (2.2, 2.3); so does its derivative aikapadika (1.14,4.1). In an attempt at explaining the meaning of aikapadika at Nirukta 1.14, Mehendale (1968a = 1978:22-30) contrasts aikapadika with prdtipadika and concludes (1978:30): 'If the meaningful units of a word into which it has been analysed explain only a single word, i.e. if the stem or the suffix so obtained can be observed in only one example, it is an aikapadika.' Noting that the word ekapada occurs only at Nirukta 2.2 and 2.3, Mehendale remarks (1978:29, note 20): 'In the former case it occurs in the sense "simple word" or primary formation as opposed to secondary formation (taddhita) and compounds {samdsa). At Nirukta 2.3, ekapada occurs in the sense "a word out of context" (prakarana-upapada-rahitdni, Durga; asahdydni avdkyasthdni ity arthah, Skanda Mahesvara).' In a review of Mehendale (1978), Bronkhorst (1984:2) criticises Mehendale's interpretation of aikapadika on the grounds that he ascribes two different meanings to it in the two single instances where it occurs, that he does the same to the related word ekapada, and that he cannot substantiate his claim that either the stem or the suffix of the enumerated words are of unique occurrence. Bronkhorst (ibid.:5) offers another solution to the aikapadika problem by looking at the objection Yaska tries to answer Nirukta 1.14. A critic has objected that if all nouns were related to verbs, they should have a form from which the meaning and the action would be understood, that is to say, words should take on the regular primary (krt) endings which their meanings require. Bronkhorst points out that 'there are words with regular krt endings.which do not have the meanings demanded by those endings and the roots to which they are attached'. On this basis he argues as follows: 'If we now understand Yaska correctly, his sentence santy alpaprayogdh krto 'py aikapadikd[h] must convey the meaning that there are words, of rare occur-

Nirvacanasastra

39

mally approached a teacher, or to someone who does not understand it, for the disdain of the ignorant towards knowledge is endless; but one should analyse to someone who has formally approached a teacher, or someone who would be fit to understand, or to the wise or an ascetic'. The statement is obviously intended to prevent, in particular, frivolous analysis based on the principle of aksaravarnasdmdnya 'similarity in syllables or in single sounds'. As will become clear later on, in this respect Yaska failed. The results achieved by these means may not correspond to what we regard as relevant linguistic facts. But they are certainly relevant semantic facts if we want to understand the texts and traditions that employed these techniques throughout the centuries and in a most striking manner. Three basic terms: artha, anvartha, and kdraka artha It has already been revealed that nirvacana analysis is concerned with semantics, with linguistic meaning. Meaning is a property in virtue of which a sentence such as 'John cooked the rice' can be used to communicate the information that John cooked the rice. The Sanskrit term for 'meaning' in this sense is artha. This term requires a few remarks. First of all, 'meaning' is itself a slippery term. We may, for example, say 'what's the meaning of all this?', 'I did not mean it', 'the meaning of this word is ...', etc. As pointed out by Quine

rence, which, though ending in krt suffixes, must be treated as grammatically unanalyzed words . . . Normally grammatical analysis (including the analysis of the Unadi sutras) helps him in his endeavour; in the case of the words damunas etc. it does not. These words are aikapadika "belonging to the unanalyzed words (ekapada)" ? Accordingly, Bronkhorst (ibid) ascribes the meaning 'unanalyzed word' ('i.e., a word where grammar does not help to reach at the meaning') to ekapada. His reasoning may be sound, but one crucial question remains: why should an unanalysed word be termed ekapadal Moreover, Bronkhorst is unfair on Mehendale when he claims Mehendale would have to offer two different translations of the word aikapadika in the two different instances where it occurs. Anyone, including Bronkhorst, would have to do that. At Nirukta 4.1 aikapadika refers to the section of the Nighantu of that name, and will invariably have to be interpreted as 'the section which deals with ekapadas' regardless of how we interpret ekapada. With regard to the latter, its occurrence at Nirukta 2.2 serves to distinguish ekapadas from secondary formations and compounds. In a similar sense ekapada occurs in the Mahdbhdsya of Patanjali (Mbh 1:6,22-3): atha vd santy ekapaddny apy avadhdrandni, 'or else, there are restrictions even in the form of single words'. Kaiyata (MbhP I:60b) explains that when one has a restriction where the word eva is explicitly used, we have a restriction which consists of two words, but when the restriction is by force of sdmarthya with no use of the word eva, then the restriction is said to be ekapada, consisting of one word; as pointed out in the Chdyd of Vaidyanatha Payagunda (ibid.), ekapada is here a bahuvrlhi compound. The term occurs also in the first vdrttika on A 1.1.63 na lumatdngasya (for which see J. Benson 1990). The vdrttika runs (Mbh 1:165,15): lumati pratisedha ekapadasvarasyopasamkhydnam, 'in the prohibition regarding [suffixes] with the element /w, an additional statement [has to be made] of the accent of a single word'. Still, Bronkhorst may well be right when he says (1984:6): 'Where and when no grammatical analysis is possible, Yaska speaks of ekapada' I do not think the problems are solved, though, by rendering it 'unanalysed word'.

40

Indian semantic analysis

(1953/1961:9) there is also a gulf between 'meaning' and 'naming' even in the case of a singular term which is genuinely the name of an object, He takes from Frege the example of 'Evening Star' and 'Morning Star'. Both of these terms name the same thing but they cannot be said to have the same meaning. There is no reason to assume that the Indians worked within a modern framework which distinguishes clearly between meaning and reference. Indeed, the term artha frequently refers to that which is designated by a term, to the thing meant. It might be useful to begin by looking at the entries on the term artha in L. Renou's Terminologie grammaticale du Sanskrit (Renou 1942/1957; 59-60): artha 'sens, signification (d'un mot, etc.); chose a exprimer, notion (not. dans paddrtha, vdkydrtha, q. v.); objet, but' P.; ifc. 'ayant le sens de' ou 'ayant pour effet, servant a': tadartha 'qui a tel sens ou tel objet, utile a cela, qui existe en vue de' P. (v. tddarthya); evamartham et kimartham M. 'dans telle intention' et-dans quelle intention?' yathdrtham M. 'selon le sens'; °arthe 'au sens de' P., ainsi dans matvarthe, q. v.; pardrtha 'qui a lieu en vue d'un su. ulterieur' ou 'en vue de qqch. d'autre' M.; uktdrtha 'dont le sens a ete deja formule' M.; v. encore, svdrtha, uttardrtha, kriydrtha, ekdrtha, dhdtvartha, anarthaka etc. 'Empechement,=nivrttV Pr. 14 48 DV. VII1 24 (cf. BR. n° 10 et Nachtrag V). He goes on to list compounds with artha as the first member and various derivatives. A second entry runs (ibid.:378): artha 'sens': l'un des elements permettant de realiser unpdda R. XVII25 (975), de differencier les nipdta XII 26 (708). Le mot est defini par a° V. Ill 1; a° est l'element essentiel selon BD. II 99. Pratiquement a° = 'phrase' dans arthddi 'initiale de phrase' Bhas II2. Ifc. 'qui a le sens de' ainsi nayatyartha toute forme 'ayant le sens de nf-' R. V 57 (373), i.e. une forme quelconque de la racine rii-. aneka°, v. ce mot; ekdrtha, v. ekasabda; arthavant 'pourvu de sens' (dit des mantra) opp. a anarthaka N. 116. Les termes de la langue parlee sont arthapurvaka V. 12, glose prayojanapu0 Uv.; les particules sont en principe arthavdcaka R. XII26 (708) et les preverbes le sont en jonction avec le nom ou le verbe XII 20 (702); arthiya ifc. N passim (dans evam° pratisedha0 upamd° etc.). Les phonemes valent sous leur forme authentique (prakrtyd) au moment ou le sens est en jeu (arthaveldydm, non 1'etude phonetique) Pratijn. 21. - Le theme nominal est arthavant GB. I 26. In his review of Renou (1942/1957) P. Thieme (1958:26) is rather critical of this: I should not finish but start with the general value 'purpose' (objet, but), from which all the others derive: the 'purpose' of [the use] of a grammatical element, a word, and a sentence amounts to what we call 'meaning' (sens, signification), and only insofar as the 'purpose' of a word may be to name a 'concept' (notion) does artha amount, in certain contexts, to 'concept'. Proceeding in this way, we have not only the theoretical advantage of having given a linguistically correct explanation of the different usages; we have also offered practical help to the non-specialist. He will now understand the different 'meanings' which must otherwise seem to him an unorganized aggregate of arbitrary values.

Nirvacanasastra

41

Thieme goes on to claim that his approach succeeds when we are dealing with certain derivatives, such as anarthaka which Thieme renders 'purposeless'. This may hit the target in a general way, but I am not at all happy with his rendering (ibid.:27) of samartha as 1) 'of the same purpose' (samdno 'rtho yasya) and 2) 'furnished with purpose' (arthena sahita-), and the important derivative sdmarthya- is just mentioned in passing in the expression sdmarthyayogdt (from Mbh 111:54,4), translated (ibid.:27) 'since it is fitting (yukta; -yogat short for -yuktatvat) that it [the teaching of Panini] be furnished with purpose'. These terms are extremely complex and do not easily lend themselves to the 'general value' of 'purpose'. It is, however, clear that Thieme has a point when it comes to establishing 'purpose' as a kind of logical semantic continuity, a constant basis for the rendering of the term artha.5* Indeed, 'purpose' is often the rendering resorted to by Sanskrit authors. For example, Kaiyata, the eleventhcentury commentator on the Mahdbhdsya, glosses anarthaka by nisprayojanam 'without purpose' (MbhP I:32b). These quotations from Renou and Thieme may suffice to illustrate that the term artha is a slippery one. In the context of nirvacana analysis this is important in as much as a nirvacana analysis is an analysis of words with regard to how they are related to things, and this is where the semantic range of the term artha becomes crucial. With the above background, let me now turn to some indigenous discussions of the term artha. An easy entry into the indigenous discussions of paddrtha, the artha of a word, is provided by certain sections of the Paspasdhnika of the Mahdbhdsya. Patanjali introduces the problem as follows (Mbh 1:6,8-11):
58

I take it that Thieme does not mean to say that artha always is to be interpreted as 'purpose' although this would be in keeping with the method that was advocated by Thieme and H. Luders in the interpretation of Vedic hymns, namely, that a certain Vedic word should be translated in one and the same way in all contexts. This has proved an untenable straitjacket. An example from Vedic is the term rtd, the ordered and regular (and thereby right, natural, and hence true) structure of the cosmic, human, and ritual course. Luders (1951:13ff. and 1959) argues that the term should be translated 'Wahrheit' (truth), partly on the basis that the word dnrta with the negative prefix an- in later language is opposed to satya- ~ 'true; truth'. But how then does one explain RV 1.105.12: rtdm arsanti sindhavah satydm tdtdna suryah, 'th floods streamed rtd, the sun spread satyd out' ? The real guardian of rtd is Varuna. Luders and Thieme strived to explain Vedic deities as developments of rather abstract concepts. Thus, Luders (ibid.) interpreted Varuna as 'Oath'. Thieme (1957) interpreted the deities Mitra and Aryaman, both related to Varuna, as 'Contract' and 'Hospitality' respectively, and Varuna as 'True speech' (based on an etymology first suggested by A. Meillet that varuna < I.E. *wer 'speak'). But it is very hard to claim that one has explained Varuna's mythological functions unless one has succeeded in explaining, for example, the antagonism between him and Indra, or the image of Varuna and the cask with the bottom up (RV 5.85.3: nictnabdram ... kdvandham). There are many aspects that cannot be explained by Varuna = True speech. Why, for example, is Varuna related to the night and Mitra to the day? Moreover, we have to ask, what did words corresponding to 'Wahrheit', 'contract', etc. mean to the Vedic Indian? A concept such as 'cosmic contract' would already change the picture quite a bit. Thieme considers it something close to a poetic and religious masterpiece that the Indians were able to colour an abstract concept so strongly. But the Vedic poets, in spite of their licentia poetica, did not have a carte blanche which allowed them to do whatever they felt like. The greater context requires something less narrow than 'true speech'. The same could be said with regard to translating artha as 'purpose'.

42

Indian semantic analysis kirn punar dkrtih paddrtha dhosvid dravyaml ubhayam ity aha I katham jndyate I ubhayathd hy dearyena sutrdni pathitdni I dkrtim paddrtham matvd jdtydkhydydm ekasmin bahuvacanam anyatarasydm ity ucyatel dravyam paddrtham matvd sarupdndm ity ekasesa drabhyate II Now, what is the artha of a word: a class (dkrti) or rather an individual thing (dravya)!59 Both, he (=Patanjali) says. How do we know? The Teacher (=Panini) has formulated rules in both ways. Considering a class {dkrti) to be the artha of a word, it is stated [by A 1.2.58] that when there is designation of a class (jdti 'class; genus; species') the plural may optionally be used with regard to one single thing. [And] considering an individual thing (dravya) to be the artha of a word, the [fact that there should be a] single remainder is taught by [A 1.2.64 which begins] sarupdndm 'of elements showing the same form'.60 Kaiyata remarks on this as follows (MbhP I:56b): tatra jdtivddina dhuh - jdtir eva sabdena pratipddyate, vyaktindm dnantydt sambandhagrahandsambhavdt I sd ca jdtih sarvavyaktisv ekdkdrapratyayadarsandd astlty avasiyate I tatra gavddayah sabdd bhinnadravyasamavetdmjdtim abhidadhatil tasydm pratitdydm taddvesdt tadavacchinnam dravyam pratiyatel suklddayah sabdd gunasamavetdm jdtim dcaksatel gune tu tatsambandhdt pratyayah, dravye sambandhisambandhdt I samjhdsabddndm apy utpattiprabhrtyd vindsdt pindasya kaumdrayauvanddyavasthdbhede 'pi sa evdyam ity abhinnapratyayanimittd ditthatvddikd jdtir vdcydl kriydsv api jdtir vidyate, saiva dhdtuvdcydl pathati - pathatah - pathanti - itydder abhinnasya pratyayasya sadbhdvdt tannimittajdtyabhyupagamahll vyaktivddinas tv dhuh - sabdasya vyaktir evdbhidheyd, jdtes tupalaksanabhdvendsrayandd dnantyddidosdnavakdsah II

This being so,61 those who hold the view that a jdti 'class; genus' [is the artha of a word] state: it is only a class that is conveyed by a word, for [otherwise] it is not possible to grasp the relation [between a word arid its artha] because there is an endless number of individual things. And it is established that there is such a thing as a jdti since we perceive the cognition of a single shape (akdra) with regard to all individuals [of the same class]. This being so, words such as gauh 'cow' denote a jdti which is inherent in different individuals. When this [jdti] has been understood [as the artha of a given word], we understand an individual thing (dravya) limited by that [jdti] since it possesses that [individual]. Words such as sukla 'white' signify a jdti which is inherent in a quality. And in the case of such a quality, cognition takes place because of [its] relation to it [= to the jdti], [while] in the case of an indiA word of caution seems necessary here. In Nyaya-Vaisesika paddrtha is the technical term for a 'category', and dravya and sdmdnya 'universality' ox jdti 'class; genus' figure among these. It could therefore be the case that Patanjah is playing around with words here in order to point out various possibilities: we are dealing with something (paddrtha) which could be dravya as well as jdti and not the one excluding the other. 60 A 1.2.64 sarupdndm ekasesa ekavibhaktau is the rule which teaches that out of a series of stems showing the same form, only one is retained before one single case ending. In other words, this is the rule which accounts for the fact that we say vrksdh 'trees', and not vrksas ca vrksas ca vrksas ca, 'a tree and a tree and a tree'. 61 The context concerns the application of the two views with regard to two panbhdsds, general rules of interpretation. This need not concern us here.
59

Nirvacanasastra

43

vidual thing (dravya), cognition takes place because of [its] relation with the relatum.62 Even words which are proper names express a jdti such as ditthatva 'Dittha-hood', [a property] grounded on an unchanging cognition so that one says 'this is that very same [person]', although there is change of [his] body in the stages of youth, adulthood, etc. from birth to death. With regard to actions too we find a jdti; indeed, it is expressed by the verbal root. Because an unchanging cognition is present from such [verbal forms] as pathati 'he recites', pathatah 'the two of them recite', [and] pathanti 'they recite', a. jdti is accepted [in the case of verbal forms] as the cause of that [unchanging cognition]. Those, however, who hold the view that a vyakti 'individual thing' [is the artha of a word] say that it is only an individual manifestation {vyakti) which is denoted by a word. But since the jdti is resorted to by implication, there is no occasion for such error as [having to assume] an innumerability [of relations]. Patanjali's initial discussion concerns directly only the context of the AstddhydyT, although Kaiyata places it within a larger frame. Later on, however, Patafijali puts his initial statement to good use in his discussion of the compound sabddrthasambandhe in the phrase siddhe sabddrthasambandhe, traditionally considered the first part of the first vdrttika of the Mahdbhdsya.63 Patafijali states (Mbh 1:7,8-10): atha kam punah paddrtham matvaisa vigrahah kriyate siddhe sabde 'rthe sambandhe cetil dkrtim ity aha I kuta etatl dkrtir hi nityd dravyam anityamll atha dravye paddrthe katham vigrahah kartavyahl siddhe sabde 'rthasambandhe cetil nityo hy arthavatdm arthair abhisambandhah II Now, with what artha of the word [siddha] in mind have we analysed [the compound] as siddhe sabde 'rthe sambandhe ca, 'when the word-form, [its] artha, and the relation [between them] are permanently established' ? He (= Patafijali) says: the dkrti 'class'. How come? Because the dkrti is permanent (nityd), [while] the dravya 'individual thing' is non-permanent. Now, if the artha of a word is the dravya, how do we then analyse [the compound]? [As:] siddhe sabde 'rthasambandhe ca, 'when the word-form and the relation with [its] artha is permanently established'. For the relation between things which possess an artha and the arthas is permanent. If the artha of the word siddha 'permanently established' is an dkrti 'class', Patafijali interprets the compound entirely as a dvandva. If, however, the artha is a dravya 'an individual thing', he keeps the last two members of the compound as a unit, itself now a tatpurusa compound: arthasambandha, 'the relation with the artha\ Kaiyata elaborates on this as follows (MbhP I:62b): dravyapakse dravyasydnityatvdd arthagrahanam sambandhavisesandrtham updttaml anitye 'rthe katham sambandhasya nityatd - id cet, yogyatdlaksanatvdt sambandhasya I tasyds ca sabddsrayatvdt — sabdasya ca nityatvdd adosah II
62

The jdti 'class; genus' (such as suklatva 'white-ness', the property of being white) is inherent in the guna 'quality' which is inherent in the dravya 'individual thing'. In this way the jdti is indirectly related to the dravya through the guna. 63 The most recent discussion of what constitutes the first vdrttikas of the Mahdbhdsya has been provided by Bronkhorst 1987; see also Wezler 1994, note 4.

44

Indian semantic analysis On the view [that a word signifies] a dravya 'individual thing', because the dravya is non-permanent, the [word] artha has been mentioned for the purpose (artha) of qualifying [the word] sambandha 'relation'. [But] if [someone asks:] when the artha is non-permanent, how can the relation [with the artha] be permanent, [we answer that this is] because the relation is defined as yogyatd 'suitability'.64 And because the word-form is the substratum of that [relation], there is no fault since the word-form is permanent. Patanjali continues his discussion as follows (Mbh 1:7,11-12; 18720): atha vd dravya eva paddrtha esa vigraho nydyyah siddhe sabde 'rthe sambandhe cetil dravyam hi nityam dkrtir anitydl katham jndyate I evam hi drsyate lokel... / dkrtdv apipaddrtha esa vigraho nydyyah siddhe sabde 'rthe sambandhe cetil nanu coktam dkrtir anityetil naitad astil nitydkrtihl katham I na kvacid uparateti krtvd sarvatroparatd bhavati dravydntarasthd tupalabhyate I Or rather, this analysis [of the compound] as siddhe sabde 'rthe sambandhe ca, 'when the word-form, [its] artha, and the relation [between them] are permanently established', is proper only when the artha of the word [siddha] is the dravya 'material substance'. For the dravya 'material substance' is permanent, [whereas] the dkrti 'shape' is non-permanent. How do we know? Because so it is seen in daily life . . . [But] this analysis [of the compound] as siddhe sabde 'rthe sambandhe ca, 'when the word-form, [its] artha, and the relation [between them] are permanently established', is proper also when the artha of the word [siddha] is the dkrti 'shape'. Weil, but it was said that the dkrti is non-permanent. That is not so; the dkrti is permanent. How? By making out that when it has come to an end in a certain place, it has not come to an end everywhere, but is [on the contrary] observed as remaining in other individual things (dravydh).

Here Patanjali considers other meanings of the terms dravya and dkrti, taking them in the completely different senses of 'material substance' and 'shape; form' respectively. He goes on to offer examples of how a material substance (dravya) such as clay or gold is permanent (nitya), while the shape (dkrti) of things made from them may vary and is therefore non-permanent (anitya). Taking dkrti in the sense of 'shape; form' he then turns the argument around, claiming that the dkrti is indeed permanent in the sense that although it comes to an end somewhere, it will still remain in other dravyas, thus incidentally resorting to the previous sense of dravya as 'individual thing'. Tired of the discussion, Patanjali concludes (Mbh 1:7,23-5): atha vd kim na etenedam nityam idam anityam itil yan nityam tarn paddrtham matvaisa vigrahah kriyate siddhe sabde 'rthe sambandhe cetil Or rather, what is the use of our saying 'this is permanent', 'this is non-permanent'? Thinking that whatever is permanent (nitya), that is the artha of the word, this analysis [of the compound] is made: 'when the word-form, [its] artha, and the relation [between them] are permanently established'.
64

That is to say, words are assumed to be permanently suited to signify, that is, refer to the things meant.

Nirvacanasastra

45

Kaiyata remarks on this in the following, illuminating manner (MbhP I:64b): yan nityam Hi I buddhipratibhdsah sabddrthah, yadd yadd sabda uccdritas taddrthakdrd buddhir upajdyate - Hi pravdhanityatvdd arthasya nityatvam ity arthahll 'Whatever is permanent'. The artha of a word is the appearance of a cognition. Whenever a word is pronounced, then a cognition appears with the shape of the artha; that is to say, the artha is permanent because it is permanent in the form of a continuous flux. Nagesa, the late seventeenth- to eighteenth-century author of the MahabhasyapradTpoddyota, or Uddyota for short, elaborates on Kaiyata's statement as follows (MbhU I:64b): yan nityam Hi I vyaktijdtydkrtTndm madhye yan nityam ity arthahll nanu sasasrngddipaddrthdndm katham nityatvam, tesdm svarupasyaivdbhdvdd ata aha — buddhipratibhdsa Hi I bdhyah paddrtho na sdbdabodhe visayah, kirn tu bauddhahl sa ca pravdhanitya Hi bhdvahll etac ca mahjusdydm vistarena nirilpitam II 'Whatever is permanent'. That is to say, whatever is permanent among the vyakti 'individual thing', the jdti 'class', or the akrti 'generic shape'. Well, but how can the artha of words such as sasasrnga 'the hare's horn' be permanent, since these have no form of their own? Therefore he (= Kaiyata) states: 'whatever appears in the mind'. In linguistically based cognition the object (visaya) is not the external artha of a word, but a mental one. And that is permanent in the form of a continuous flux. This is what Kaiyata means to say. And this [topic] has been discussed at length in the Manjusd.65 From Nagesa's last remark it is evident that this brief presentation is not in any way an exhaustive discussion of the indigenous elaborations on the term artha. But it has sufficiently established that the distinction between meaning and thing meant is extremely blurred, and that the discussion revolves around three possibilities: that the artha of a word is a class (jdti, dkrti), a shape or form that appears in the mind (akrti, dkdra), or an individual thing (dravya, vyakti). It is also clear that in vydkarana questions of meaning become questions of reference. The gulf between 'meaning' and 'naming' has narrowed considerably, and this is one of the reasons why it is possible to arrive at the artha or 'meaning' of a term through that term itself by means of nirvacana analysis. Now, it is also clear that the term artha is used in a more abstract way than simply referring to the thing meant. As far as I know, it is in the Nirukta itself (Nir 1.3) that we find the earliest extant discussion on whether certain linguistic elements express an artha 'meaning' or 'purpose' (arthdn nirdhuh). The
65

For example m Nagesa's Vaiydkaranasiddhdntalaghumanjusd {bauddharthamrupana, 203-369).

46

Indian semantic analysis

context is the meaningfulness of upasargas - preverbs or prepositions. This serves to demonstrate that the term artha implies more than just 'thing' or 'referent', since the meaning of preverbs surely is rather abstract. The term artha is also used to denote meaning in a broader sense, more specifically with regard to the meaning of a corpus of texts. Sayana, for example, concluded his exposition of nirvacanasdstra in the following manner (Say p. 29): tasmdt veddrthdvabodhdyopayuktam niruktam, 'therefore nirukta is useful for grasping the meaning {artha) of the Veda'. Similarly, the commentator Durga remarks with regard to the Nighantu (D 1:30,3-4): sa ca rsibhir mantrdrthaparijndndyoddharanabhutah paficddhydyisdstrasamgrahabhdvenaikasminn dmndye granthikrta ity arthah, 'that is to say, in order that [we shall get] knowledge of the meaning (artha) of the Vedic verses, the Rsis have also made it [=the sam-dmndya] into a text that serves as an example in the form of a compendium of a Sastra which consists of five parts in one single [ekasmin-sam, cf. Nir 1.3] dmndyah (repetition within certain limits)'. Clearly it is the meaning of the Veda Yaska has in mind when he quotes the following verse (Nir 1.18): yad grhitam avijndtam nigadenaiva sabdyate I anagndv iva suskaidho na taj jvalati karhicitll What has been taken [from the teacher's mouth] but not understood [and] is uttered by mere recitation, that never flares up, like dry firewood on a non-fire. The verse is quoted also by Patanjali in the Paspasdhnika of the Mahdbhdsya (Mbh 1:2,15—16),66 a clear example of how Patanjali shows Yaska respect in this introductory chapter. Yaska also comments on a Vedic verse (RV 10.71.4) which makes his attitude quite clear (Nir 1.19): utd tvah pdsyan nd dadarsa vacant utd tvah srnvdn nd srnoty endm I utd tvasmai tanvdm vi sasre jdyeva pdtya usatf suvasdhll And many a one, [although] seeing, does not see Speech; and many a one, [although] hearing, does not hear Her; and to many a one She spreads out [Her] body, like a wife wearing clean clothes, having desire for [her] husband. This verse too is quoted by Patanjali (Mbh 1:4,2-3). His interpretation agrees more or less with Yaska's, except for one significant point: to Yaska the wife who uncovers herself uncovers knowledge, the illumination of meaning (artha) (svam dtmdnam vivrnute jndnam prakdsanam arthasya), and the verse is a praise of one who knows the meaning (arthajnaprasamsd); to Patanjali the verse says that speech uncovers itself to someone who knows speech (vdg vdgvide svdtmdnam vivrnute), that is to say, to the grammarian. Yaska (Nir 1.20) goes on to quote the subsequent verse (RV 10.71.5), and remarks: artham vdcah puspaphalam aha, 'he declares the meaning of speech to be its fruit and flower'.
66

Reading adhitam 'studied' for grhitam.

Nirvacanasastra

47

Finally, the following passage from the Mahdbhdsya may serve to illustrate the intricacies of the term artha (Mbh 1:363,24-8): apravrttih khalv apy arthadesanasya I bahavo hi sabdd yesdm arthd na vijfidyante I jarbhdnturphdritull antarena khalv api sabdaprayogam bahavo Wthd gamyante 'ksinikocaih pdnivihdrais call na khalv api mrjhdtasydrthasydnvdkhydne kim cid api prayojanam asti I yo hi bruydt purastdd dditya udeti pascdd astam eti madhuro gudah katukam srngaberam iti kim tena krtam sydtll There will, in fact, be no attempt at instruction in meaning {artha). For there are many words whose meanings are unknown, such as jarbhdn turphdntu.61 In fact, many meanings are understood without the use of words, by closing the eyes or by doing things with the hands. Indeed, there is no purpose whatsoever in explaining meaning which has [already] been understood. For he who says that the sun rises in the east and sets in the west, that sugar is sweet, [or] that ginger is pungent, what is accomplished by him'? Let these considerations suffice to establish that the term artha is as complex an entity as its English counterpart 'meaning' which I shall use to translate it. Let it also be clear that we shall first and foremost be dealing with linguistic meaning in the sense that we deal with the meaning of singular and general terms. In that context questions of meaning are primarily questions of reference. In order to account also for the broader and the more abstract senses of artha, however, I shall confine myself to saying that words signify. That this is a sensible safeguarding is also brought out by the related term anvartha which plays a central role in nirvacana analysis and to which I shall now turn. anvartha In the introduction to his Rgvedabhdsya Sayana remarks on the lexicographical work Nighantu in the following way (Say p. 28): pancddhydyarupakdndatraydtmake etasmin granthe paranirapeksatayd paddrthasyoktatvdt tasya granthasya niruktatvam, 'the property of being nir-ukta (stated independently) pertains to this text [= the Nighantu], since in this text, which consists of three sections forming five chapters, the meaning (artha) of words has been stated (-ukta) independently (nir-) of other [words]'. Sayana gives a similar analysis of the name of Yaska's Nirukta (ibid.:29): tad api niruktam ity ucyatel ekaikasya padasya sambhdvitd avayavdrthds tatra nihsesenocyante iti vyutpatteh, 'according to the analysis: "the presumptive (sambhdvitdh) meanings of the constituent parts (avayavdrthdh) of each individual word68 are stated (ucyante: -ukta) fully (nihsesena: nir-) in it", this too is called Nirukta\ Both these analyses are worked out with an eye to the nature of what the terms denote. An analysis of this kind as well as a term treated in this manner are
67 68

RV 10.106.6 Or, possibly, ekaikasya could mean 'of one word after the other' with reference to the list of words constituting the Nighantu

48

Indian semantic analysis

both said to be anvartha: 'in accordance with that which it signifies', or 'in accordance with the meaning and nature of that which it denotes', or, even, 'purposive'. So when a term or expression is anvartha, it is inherently significant or purposive so that there is a direct semantic link between the term and the nature of that which it denotes. Without entering into the complexities pertaining to singular reference, it is clear that the name 'John' does not in itself tell us anything about the man so named. It is just a label by which he is called. But a term such as 'Bhairava' is actually descriptive. It conveys the nature of that which it denotes in the sense that we may learn about Bhairava's nature from the name 'Bhairava' itself. So such terms are descriptive terms as well as names. It should be noted that this feature is in no way limited to proper names within the context of Indian linguistic speculation. It applies to all nominal words. The term 'Bhairava' is thus treated as a descriptive name in the sense that the term itself contains the necessary elements to provide the uniquely true description. It is thus similar to a name such as 'Jack the Ripper'. In ordinary English very few names can be considered descriptive names in this sense. Within the Indian mode of thinking, however, any name, any noun, may in principle be considered a descriptive name in as much as it is susceptible of nirvacana analysis, semantic analysis on the basis of the term itself. The explanatory expression provided through such analysis is thus linked up with the linguistic item itself; any uniquely true description not so linked would not do. In other words, it is the very power of nirvacana analysis to give for any term a unique description of its referent through the term itself, thus forcing the term to reveal its semantic content. This is fundamental to the way in which a nirvacana analysis works. Linguistic entities are treated as inherently significant, and this significance can be arrived at through the linguistic element itself because the linguistic element is anvartha. The term anvartha is well known also from vydkarana. On several occasions in the Mahdbhdsya Patafijali states (Mbh 1:206,24; 11:366,25; 111:409,2): anvartham khalv api nirvacanam, 'the nirvacana 'does indeed correspond to that which [the term] signifies'. Patanjali also repeatedly states that the purpose of making a long technical term in grammar, a mahati samjnd, is that it should be recognised as an anvarthasamjhd, 'a technical term which is in accordance with that which it signifies'.69 A single example
69

Kiparsky (1979:210) expresses a different opinion when he concludes: 'In any case, such outside meanings have at most a mnemonic value and are strictly irrelevant within the theory.' He adds in a note (ibid.: note 1): 'This is against the opinion of Patanjali, who often seeks a special significance in "long" terms.' This is not only against the opinion of Patanjali, but against the opinion of the entire vydkarana tradition. Kiparsky is probably right, though, in claiming that Panini himself coined or viewed a certain term as an anvarthasamjnd. As pointed out by Cardona (1977:340), if one were to attribute such a view to Panini, one would also have to consider one rule vacuous (A 1.1.41 avyayibhdvas ca [avyayam 37]). Cardona goes on to say (ibid.):

Nirvacanasdstra

49

will suffice.70 A 1.4.23 kdrake introduces the technical term kdraka, lit. 'accomplishes doer', as the technical term for participants in actions. This term plays a considerable role with regard to nirvacana analysis, and as such I shall discuss it in some detail below. On A 1.4.23 Patanjali remarks (Mbh 1:324,7-10): kdraka iti mahatT samjhd kriyatel samjhd ca ndma yato na laghvtyahl kuta etatl laghvartham hi samjhdkaranam I tatra mahatydh samjhdydh karana etat prayojanam anvarthasamjnd yathd vijhdyeta I karotiti kdrakam iti I [The term] kdraka is made as a long technical term. But a technical term {samjhd) is that than which nothing is shorter. How come? Because technical names are made for the sake of brevity.71 In that case, the purpose in making a long technical term is this that it should be recognised as a technical term in accordance with that which it signifies (anvarthasamjnd), [according to the analysis]: 'it accomplishes [something], hence [it is called] kdraka\ Yaska too analyses technical terms in the manner of an anvarthasamjnd. His analysis of the technical term nipdta 'particle' as presented by Sayana will serve to illustrate this. Sayana states (Say p. 29): tatra hi'catvdri padajdtdni ndmdkhydte copasarganipdtds cay iti pratijhdya 'uccdvacesv arthesu nipatantV iti nipdtasvarupam nirucya evam uddhrtam - 'neti pratisedhdrthlyo bhdsdydm ubhayam anvadhydyam nendram devam amamsateti pratisedhdrthlyaK iti; 'durmaddso na surdydm ity upamdrthiyaK (Niru. 1.1; 4) iti cat tac ca loke kevalapratisedhdrthiyasydpi nakdrasya vede pratisedhopamdlaksanobhaydrthoddharanam asmin granthe 'vagamyate I For therein, having asserted that 'there are four classes of words: nouns, verbs, preverbs and particles' (Nir 1.1) and having proclaimed (nir-ucya) the real nature of [the word and that which is referred to by the word] 'particle' (nipdta) through [the analysis:] 'they occur (nipatanti)12 in all sorts of meanings' (Nir 1.4),73 he exemplifies in Footnote 69 (cont.) Though this is a small point, its repercussions are not. Let us agree that Panini did indeed take such terms from earlier grammatical work. Since Panini also formulated rule 1.1.41, it follows that he did not consider the term avyayibhdva significant in the sense that from this term alone one knows that an avyayibhava is an avyaya. On the contrary, this term denotes any compound formed by a rule under the heading of 2.1.5 (avyayibhdvah). Now, we also knowfromPatanjali that some grammarians defined classes of compounds in terms of their meanings [...] and that Panini did not. It would be reasonable - though unprovable - to say that the two things go together: the use of avyayibhdva as an anvarthasamjna and the semantic scheme for classing compounds. One would then have to say that Panini represents a break with such a system, if indeed this is all pre-Paninian, and a move towards formalism, towards defining classes relative to operations proper to them. This, however, does not have any bearing on the fact that the grammarians after Panini exploited the fact that major technical terms are considered significant. 70 Other examples are found Mbh 1:81,26-9; 88,27-89,3; 96,11-13; 215,7-11; 346,16-18; 378,17-19; 11:3,5-8; 76,6-10. 71 As pointed out by Cardona (1969:29, note 76), the Padamahjarl on the Kdsika at A 2.1.5 (KasP II: IB) even72claims that a monosyllabic technical term should be coined: ekdksard samjnd kdryd. Lit.: 'they fall into'; glossed vartante by Skanda-Mahesvara (SM 1:46,3). 73 This nirvacana is quoted by Sayana also at RV 1.124.12, as noted by Sarup (1927:30, note 3).

50

Indian semantic analysis the following manner: [the particle] na has the sense of negation in colloquial language, [but] both [senses, that is, negation and comparison,] according to that which is to be studied74 [= Veda]; [thus, in] 'they did not consider Indra a god' (RV 10.86.1) it has the sense of negation, but [in] 'like people drunk on wine' (RV 8.2.12) it has the sense of comparison.75 And this is an instance met with in this text [=the Niruktd] of both senses of the syllable na, indicating negation and comparison in Veda while in colloquial language it has the sense of negation alone.76

It appears to me, then, that just like the meaning of declarative sentences, the meaning of an anvarthasamjnd or the outcome of an acceptable nirvacana analysis of any singular or general term, is intimately connected with whether it is true or false, the sentence as a sentence, the term as to whether it is an adequate name. The sentence 'John cooked the rice' is true if and because John cooked the rice. The nipdtas are nipdtas if and because they nipatanti 'occur [in all sorts of meanings]'. Several of the text-passages I shall turn to later on deal in some detail with the notion that terms are anvartha. karaka Before I turn to text-passages which illustrate the workings of nirvacana analysis in practice, I find it necessary to address one more issue fundamental to nirvacanasdstra. This is the notion of karaka mentioned above. In Paninian grammar the term karaka (lit.: 'doer; accomplished) applies to
74

75 76

This is how I interpret the word anvadhydyam (Nir 1.4), that is, as karmavyutppttika, as a noun liable to an analysis as the object of the verbal action involved. Skanda-Mahesvara, on the other hand, prefer to interpret as karanavyutpattika, as a noun liable to an analysis as the instrument j of the verbal action involved (SM 1:46,6): anvadhydyam ity adhydyasabdah - 'adhydyanydyodydve'ti karanddhikardnayor nip The item adhydya in anvadhydyam is ready-made {nipdtyate) in the senses of instrument or location in accordance with [A 3.3.122:] adhydyanydyodydvasamhdrd[dhdrdvdyd]sca. This rule teaches that also the ready-made or irregular forms (nipdtana) adhydya 'chapter', nydya 'rule; maxim', udydva 'mixture', samhdra 'destruction', and [according to the Kdsikd and the editions of Vasu (1891) and Katre (1987) also] ddhdra 'support' and dvdya 'cloth-mill' (so Katre) are derived with the krt-snffix GHaN to form nouns in the masculine in the sense of instrument or locus. The example offered by the Kdsikd is adhlyate 'sniinn ity adhydyah '[something] is studied in it, thus adhydya\ in the sense of 'chapter' or 'lesson', but from the continuation of the commentary it is clear that Skanda-Mahesvara have the instrumental alternative in mind: adhiyate 'nena '[something] is studied by means of it' (ibid.^ 6-8): vedas ca yady apy adhyayanasya karma, tathdpi karmanah kvacit karanatvena viv drsyate, odanena bhunkta itil ato 'tra karanatvena vivaksite vede varttatel But even though Veda is the object of the act of studying, one nevertheless meets with a desire to speak of the object as an instrument, as in 'he eats by means of rice'. Thus it occurs here in the sense of Veda which there is a wish to express by [its] property © being an instruf ment [of study]. Cf. Nirukta 1.4. In this passage it would certainly be possible to translate artha as 'purpose'; for example, a nipdta such as na serves the purpose of negation.

Nirvacanasastra

51

direct participants in actions.77 Such a participant is a means of realising the action, a sadhana™ and every participant is assigned to one of a set of six kdraka categories. In Panini's grammar these are defined in the following order:79 apdddna 'stable point when there is movement away', sampraddna 'recipient; indirect goal',80 karana 'instrument', adhikarana 'locus', karman 'object; goal', and kartr 'agent'. Moreover, a subcategory of agent is defined, namely hetu, the causal agent. The abstract syntactic level at which kdrakas are introduced in the grammar serves to mediate between the levels of semantics and morphology. By this device Panini is able to account for the relationship between possible semantic choices on the side of the speaker and some basic features of Sanskrit syntax and morphology. In practice, the advantage is that at the abstract syntactic level of kdrakas these structures can be treated as identical and thus share operations. For example, a passive sentence is in no way derived from an active sentence. Both are simply alternative manifestations of what at the underlying abstract kdraka level is represented as the same idea. This means that in the process of derivation the starting point is semantics, semantics here including such features as past, present, and future time, active, passive or stative voice, and the participants in actions. Such participants in actions are, at the level of kdraka syntax, assigned to particular karate-categories. In a similar way the set of lakdras, a set of ten suffixes marked with an L, are introduced after verbal roots at the same abstract level. At this level all verbal endings, except for the bhdva ones, can be said to signify agents and objects in relation t o activities and can thus be treated according to what they at this level have in common. By A 3.4.69 lah karmani ca bhdve cdkarmakebhyah an L-suffix is added to a verbal root to denote - in addition to the agent (kartari, A 3.4.67) - the object, or, in the case of verbal roots which are objectless (akarmaka, that is, intransitive verbs), the mere activity expressed by the verbal root (bhdva, lit.: 'being'). Moreover, these Lsuffixes serve to mark tenses. For example, A 3.2.110 lun teaches that the aorist L-suffix IUN is added after a verbal root (A 3.1.91 dhdtoh) when the action refers to the general past (A 3.2.84 bhute). These abstract L-suffixes are subsequently replaced by finite verbal endings or by participial suffixes. When the relevant semantic choices have been considered on the side of the speaker, the abstract syntactic level of kdrakas and lakdras is sorted out. The correct distribution of case endings and finite verbal endings is then accounted for in the syntax of a Sanskrit sentence by means of operational rules. By letting sentences be subject to shared operations at the abstract syntactic level,
77 78 79 80

On kdraka in Paninian grammar, see, for example, G. Cardona 1967; 1974; 1976:215-24; 1988:160-2; S.D. Joshi and J.A.F. Roodbergen 1975; P. Kiparsky 1982. I analyse this term as sadhyate 'nena, '[something] is realised/accomplished through it'. Rules A 1.4.24-55. As G. Cardona hasrightlypointed out (1976:340, note 273; 1988:168-9), the terms apdddna and sampraddna do not easily lend themselves to any one translation. The apdddna category, for example, includes a stable point when there is movement away, an object of fear, etc.

52

Indian semantic analysis

Panini is able to account for the relationship between semantics and case endings, between sentences with finite verbs in the active, passive or impersonal passive (stative) voice, and between sentences containing finite verbs and nominal sentences. Moreover, he succeeds in accounting for syntactic problems pertaining to such areas as control81 and ellipsis.82 An example may serve to clarify this. Consider the following sentences: puruso vrksam chinatti, 'the man is cutting the tree', and purusena vrksas chidyate, 'the tree is being cut by the man'. To end up at one or the other of these levels, derivation proceeds through three stages. My intention here is simply to give the skeleton of the procedure, so I do not list the rules which map level (1) onto level (2) or level (2) onto level (3), nor do I list the conditions which affect these rules at level (2) and level (3). (1) Semantic level introducing the participants in the action of cutting: DUriiSa^svatantra' l^e m(*ependent participant) _j_ yy.^^{ipsitatama, that which is most desired by the agent) i -\Jrhid + JATivartamdna, present time) (2) Abstract syntactic level; participants assigned to karaka categories: purusa(kartr) + vrksa(karman) + ^Ichid+lApkartan) or: purusa{kartr) + vrksa^karman) + ^lchid+lApkarmam) (3) Surface level of morphology: purusa + sU+ vrksa + am + A chid+SaP + UP — puruso vrksam > chinattP or: purusa + Td + vrksa + sU+ ^Ichid+yaK + te ->purusena vrksas chidyate Thus, looking back at given semantic conditions, kdrakas and la-kdras serve to derive correct Sanskrit sentences by sorting out an abstract syntax on the basis of which operational rules distribute the proper morphology. It should be noted, however, that the above two sentences could equally well
81

That is, the problem of accounting for the implied agents and objects of non-finite verbal form such as absolutives and infinitives Consider, for example, the shift of agent lathe case of absolutives as illustrated by the two English sentences 'Having arrived in the village, John cooked the rice', and 'Having arrived in the village, the rice was cooked by John' 82 That is, the problems linguists face as a result of the fact that words may be elided in sentences. The Piccadilly Line' is a statement one can only imagine would occur as the reply to a question such as 'Which line do I take to get to South Kensington9', and as such an incomplete sentence for To get to South Kensington you take the Piccadilly Line.' Panini treats constructions such as edhebhyo vrajati, 'he goes for firewood', as elliptical, equal to edhdn dhartum vrajati, 'he goes to fetch firewood' For recent discussions of case, control, and ellipsis in Pamman syntactic theory, see P. Kiparsky 1982 and M.M. Deshpande 1985a. 83 We may at this stage speak of a fourth level of phonetics at which the correct sandhi rules apply, but that need not concern us here

Nirvacanasastra

53

appear in the form of corresponding nominal expressions. With an agent noun and a genitive construction we get puruso vrksasya cchettd, 'the man [is] a cutter of the tree', and with a past participle in a passive construction and the agent in the instrumental case we get purusena vrksas chidyamdnah, 'the tree [is] being cut by the man'.84 Disregarding details of morphology, the derivation procedures would be the same. But kdrakas do not pertain to the derivation of sentences alone. Any verbal noun derived by a &rr-suffix (a primary suffix) is considered to denote either a participant in an action and thus assigned to one or the other of the kdraka categories, or the mere activity expressed by the verbal root and thus assigned to the category of bhdva 'being; state of action'. In other words, a certain krtsuffix is added to a verbal root when a particular kdraka, participant in the action denoted by the verbal root, is to be denoted, or when the mere verbal process or state of being {bhdva) is to be denoted. By way of example, the suffix LyuT {-ana with guna and presuffixal accent) is introduced to form neuter action nouns {bhdve) (by A 3.3.115 lyut ca), but also to derive nouns expressing the instrument or the locus of the action expressed by the root (by A 3.3.117 karanddhikaranayos ca). This means that when analysing a particular word alternative interpretations are often possible. For example, the word yajana can be interpreted as bhdvasddhana, that is to say, as an action noun (with LyuTby A 3.3.115 lyut ca) denoting the mere activity of sacrificing. But it could alternatively be interpreted as adhikaranasddhana, that is to say, as a noun denoting a locus, 'place of sacrifice'; or even - theoretically - as karanasddhana, as a noun denoting a means of sacrificing (the two latter formed with LyuTby A 3.3.117 karanddhikaranayos ca which teaches that this suffix is added also to denote the instrument and the locus). This, roughly, is how kdrakas work in the Paninian tradition. What this brings out is crucial: not only are kdrakas not limited to the derivation of sentences, but any verbal noun derived by a fc/t-suffix (a primary suffix) is treated as if it were a sentence in its own right. To Yaska and within the realm of nirvacana analysis this is the case not only for primary nouns but for all nouns. Let me return briefly to the verse from the Atharvaveda (3.13.2) quoted earlier. The second half of the verse runs: tad dpnod indro vo yatis tdsmdd apo dnu sthana, 'then Indra got hold of you as you flowed, hence you are Waters thereafter'. By this formulation the author has formulated an analysis of dpah as karmasddhana, as a noun denoting an object, karman. The same holds good with regard to the more technical formulation in which this reappears in the Satapathabrdhmana (6.1.1.9): yad dpnot tasmdd dpah, 'that he got hold of [them], on account of that: dpah\ The same analysis has found an even more technical formulation in the Nirukta (9.26): dpah dpnoteh. Here the nirvacana is expressed in a way which is not immediately illuminating with regard to
84

For an overview of such core paradigms of Sanskrit syntax, see P. Kiparsky 1982 2-3

54

Indian semantic analysis

kdraka analysis. The reader will have to make his own interpretation. It seems to me to be embedded even within? a condensed version of a more elaborate expression. If an analysis does not bring out an obvious interpretation with regard to kdraka, the commentators invariably supply one or, quite frequently, several. This issue will be addressed at length further on. However, Yaska frequently formulates a nirvacana in a more elaborate manner which leaves no doubt with regard to kdraka. A few examples will suffice to illustrate the point. We meet with a clear apdddnasddhana, an analysis as a noun denoting that from which there is movement away, of the word dhanvan (Nir 5.5): dhanvdntariksaml dhanvanty asmdd dpah, 'dhanvan, i.e. antariksa (the intermediate space between heaven and earth): [something] flows (dhanvanti) from it, [namely] water'. The word ahar 'day' is interpreted as adhikaranasddhana, as a noun denoting a locus (in this case as the time in which something is done) (Nir 2.20): ahah kasmdtl updharanty asmin karmdni, 'why85 ahar! [because] one carries out (updharanti) [something] in the course of it, [namely] actions'. The word gnsma 'summer' is also given an interpretation as adhikaranasddhana (Nir 4.27): grismo grasyante 'smin rasdh, 'grismah: [something] is devoured (grasyante) during it, [namely] juices'. Whether or not a nirvacana is explicit with regard to kdraka analysis, one of its crucial features is nevertheless that the term analysed is linked to some action, and that ideally it is also specified in what way that which the term signifies participates in this action, in exactly the same way as nouns formed by Art-suffixes are assigned to kdraka categories, or participants of actions in the derivation of sentences in Panimya grammar. With this background, let me now consider how this framework is applied in practice.
85

See Kahrs 1983 for this interpretation of kasmat in the Nirukta

3 Praxis: Saiva Kashmir

In the Appendices to his translation of Kalhana's Rajatarangini, Marc Aurel Stein discusses the texts known as mdhdtmyas and connected to all the more important tirthas of Kashmir. These mdhdtmyas are primarily handbooks for the purohitas of the particular tirthas, that is to say, for the priests who receive the pilgrims. The purpose of a mdhdtmya is to ascertain and expound the sacred nature of the tirtha and the spiritual rewards for visiting it. With few exceptions the mdhdtmyas claim to be extracted from Puranas or Puranic collections, but Stein offers some quite revealing insights here. Concerning the Purohitas, he has the following to say (Stein 190011:380): These local priests known in Kasmlr as thdrfpqt1 (Skr. sthdnapati), are as a rule quite as ignorant and grasping as their confreres, the Pujarls, BhojkTs, etc., of India proper. They are held deservedly in very low estimation by the rest of the Brahman community. That their condition was more or less the same in earlier times too, though their influence and numbers were probably greater, can be safely concluded from more than one ironical allusion of Kalhana. These are the people to whose keeping the Mahatmya texts have always been entrusted. Stein (ibid.) points out how not only the route of the pilgrimage, but the very site of the tirtha had sometimes been changed in times relatively recent to his visit. Obviously, the choice of these subsidiary cult-places must have depended on local considerations from the very beginning. When these considerations changed, variations in the pilgrimage routes inevitably followed. Stein explains how this affected the mdhdtmya texts (ibid.:381): To bring the text of the Mahatmya into accord with these successive changes was a task which devolved upon the local Purohitas, and the texts we have discussed above bear only too manifestly the traces of their handiwork. Sound knowledge of Sanskrit and literary culture are likely to have been always as foreign to this class of men as they are at present. When it became necessary for them to introduce the names of new localities into the text of the Mahatmya, there was every risk of these names being shown, not in their genuine old forms, but in hybrid adaptations of their modern Kasmlrl equivalents. This risk naturally increased when Sanskrit ceased to be the official language of Kasmir, and the knowledge of the old local names became gradually lost even to those maintaining scholarly traditions in the country.

56

Indian semantic analysis

As 'another potent cause' that seems to have co-operated in this vitiation of the mdhdtmyas' local nomenclature, Stein points to 'popular etymology', that is to say, the imprints of nirvacanasdstra. He says (ibid.)'. We have already referred to the tendency displayed throughout these tracts of making the names of localities, rivers, springs, etc., the starting point of legendary anecdotes. For men of such very scant knowledge of Sanskrit as the Thanapat1s invariably are, it was naturally far easier to explain such etymological stories when they were based on the modern local names. It is undoubtedly this reason which has, e.g., led the author of the present Harumukutagangamahatmya to substitute the name KarankanadT for the old Kanakavahinl. By the latter name the stream coming from the Harumukuta lakes is designated in all our old texts, as explained in my note on Rajat. i. 149, 150. By turning Kdnkanai, the modern derivative of this old name, into KarankanadT, the 'skeleton-stream', the compiler of the Mahatmya got an occasion to treat his readers to a legend likely to appeal to their imagination. The river is supposed to have received this appellation because Garuda had dropped at its Samgama with the Sindhu the skeleton (karanka) of the Rsi Dadhlci which Indra before had used as his weapon, etc.1 This story, it is true, is wholly unknown to the Nllamata or any other old text. But, on the other hand, it has the great merit of being easily explained and proved to any Kasmirl pilgrim. The latter cannot fail to realize the manifest connection between Karanka and his familiar J^ranz, 'skeleton'. As Stein points out, it would be easy to multiply examples showing the strange vicissitudes to which old topographical names are exposed at the hands of the local Purohita. Let the previous one suffice to illustrate how nirvacanasdstra was put to good use in religious discourse as recently as the late nineteenth century.2' This is a good example of how meaning is created by the community, by
1 2

As pointed out by Stein (ibid., note 150), the story is spun out at great length in Patala iii. of the Harumukutagangamahatmya, No. 221. I find it irresistible, however, to add one further example of how a text is not a given entity but liable to innovation, change, interpretation, and tampering. Stein (ibid.: note 153) tells the following story: 'I am glad that chance gave me an opportunity of gaining some personal experience of the manner in which Mahatmyas are occasionally produced. Some ten years ago the Purohitas or Bachbattas of the Ganapatyar quarter in Srlnagar recovered an ancient Linga from a mosque and began to erect a small shrine for it near the river Ghat of MaPyar. Guided by a local tradition which, as far as I can judge, may be genuine, they believed this to have been the site of the shrine of Siva Vardhamanesa, mentioned already in the Rajatarangim (see note ii. 123). The Linga was re-consecrated accordingly by this name. In 1891 when examining old sites in this part of the city, I also visited the temple of Vardhamanesa then under construction. The interest I showed in the old Linga and the tradition regarding it, coupled with an appropriate Daksina, soon secured me the confidence of the head-Purohita of the little shrine. Pandit T. R., a man more intelligent than the average of his fraternity, was not slow to confess to me that the Mahatmya of the Tirtha in spe was as yet under preparation. Some weeks later, when in camp near Srlnagar, I received the visit of my Purohita from Vardhamanesa's shrine. He brought me the draft of the new Mahatmya and asked my assistance in revising it. I found it consisting chiefly of extracts from the Vitastamahatmya. The passages dealing with Vardhamanesa and the neighbouring Tirthas within the city had been suitably interlarded with laudatory verses in the usual Mahatmya style culled from other texts. The vested interests of other local shrines had received due recognition by being included in the Ydtrd of Vardhamanesa. I did what I could to indicate the genuine old names of these localities. This quasi-antiquarian co-operation does not seem to have detracted from the popularity of the new Mahatmya among the Bachbattas of Ganapatyar.'

Praxis: Saiva Kashmir

57

people interpreting, and of how the capacity of defining meaning and thus the contents of thoughts rests not only on somebody's ability to be semantically creative, but also on the same person's social position to be so. It also illustrates how there is a tradition in exegesis. However, to illustrate the role played by nirvacana analysis in exegesis and metaphysics, I shall not stay with Stein's Purohitas but turn instead to the learned Sanskrit literature of Saiva Kashmir around the turn of the millennium, a literature to which similar practices, albeit more sophisticated, are not unknown. More specifically I shall turn to the exegetical activity focused on Tantric Saiva scriptures from the tenth to the thirteenth centuries CE. The reason for this choice is the intensely active exegesis the scriptural revelations known as Tantras were subject to during this most articulate and formative period of Tantric Saivism. Moreover, these Kashmirian systems of ritual and exegesis of ritual became the backbone of Tantric orthodoxy in southern India from the eleventh century onwards; from there they spread throughout India. In a narrow sense, the term tantra is used with reference to a set of rituals or basic instructions and to the texts which teach them. Those who act in accordance with these practices and the scriptures related to them are known as Tantric (tdntrika) as opposed to those who hold as authoritative Vedic scriptures (vaidika). For a general introduction to Saivism and the Tantric traditions, I refer to Alexis Sanderson (1988/90).3 The Saivas were not the only Tantrics, and certainly not the only ones to make use of nirvacana analysis in their exegesis. One could instead have turned to the Vaisnava Tantric tradition of the Pancaratra, in particular to the Ahirbudhnyasamhitd of the Pancardtrdgama, a text composed in south India in the eleventh century.4 The great prestige of the non-dualism propagated by the Kashmirian Saivism is evident from the fact that it was incorporated into the framework of this text, but the reason it makes a special case for nirvacana analysis is because the Pancaratra tradition felt a need to put on a Vedic image when it moved south. At first it incorporated Taittirlya material in their ritual, then it went further in colonising pan-Brahmanical material. The relevant sections of the Ahirbudhnyasamhitd put nirvacana analysis to use in careful analysis of Vedic mantras; an absolutely universal mantra such as the Gayatrl is, for example, forced into a strictly Vaisnava meaning.5 The case of Bhairava: preliminaries As my prime focus of study I have chosen the name of the Tantric deity Bhairava. The names of Tantric deities are extremely important since the ritual
3 5

I shall use the term Tantric in a general way to cover both Tantric and Kaula forms of ritual; see A. Sanderson 1990, note 1. 4 See Sanderson 1990:34, note 16 for this dating. ABS, adhyaya 57 ff., the pahcahotrmantra (58), the purusasukta (59); other examples are the visnumantras in adhyaya 52 and the jitantastotras in adhyaya 53 (where each word is analyse metaphysically), or adhyaya 54 dealing with the cult of the sudarsanacakra, an aspect of Narasimha.

58

Indian semantic analysis

of these Tantric cults is mainly concerned with their evocation by means of mantras, sequences of power-laden syllables. These mantras too are therefore at the core of religious practice. There is one more reason for introducing a proper name as an example: utterances containing proper names are dependent upon the existence and coherence of a general practice of reference.6 In short, Bhairava is a male deity known from south Indian inscriptions dating from around 400 CE as the object of an exclusive theistic cult. Accordingly, his status as a major deity is established from at least that time, but it probably dates back to a considerably earlier period. Among later evidence we find that iconographic features of Bhairava are mentioned in a Tevdram hymn, the most famous portion of the Tirumurai, a collection of Tamil Saiva hymns composed from the sixth century onwards. The Saiva Kailasanatha temple at Ellora (founded c. 775) contains a Bhairava shrine. Bhairava is clearly associated with Rudra, and is seen to hold a strong position among the Kapalika (from kapdla 'skull') cult of power through impurity associated with the Saiva culture of the cremation grounds.7 This non-dualistic Kapalika Saivism was based scripturally in the Bhairavatantras,8 texts which have the form of Bhairava's teachings to the Goddess BhairavT (Devi), but embraced also the Bhutatantras, concerned with control of various powers through the worship of Bhairava, and texts concerned with the worship of Kali, the cult of whom was later to be considered the highest form of Saivism with the incorporation of the goddess-worship adapted from the system known as the Krama. The term Krama refers to a number of esoteric cults of intense Kali worship in a sequence {krama) of sets of divine Powers; the Krama conceives of ritual as a means of liberating insight, and the sequences consist in the worship of goddesses which represent the cycle of cognition frotti initial to final voidness.9 In Kashmir towards the end of the first millennium a particular Bhairava cult had become what seems to have been the norm. But the deity of this cult, called Svacchandabhairava 'Autonomous Bhairava', was no longer the bloodthirsty deity of the cremation grounds. We find him here as a slightly domesticated and modified version of the wilder Bhairava, and worshipped together with his female consort Aghoresvarl. At the time of Abhinavagupta (fl.c. 975-1025 CE)10 this cult of Svacchandabhairava with his female consort was, as it has remained, the basic Saivism in Kashmir. To the one side of this cult was the du&listic Saiva Siddhanta, originally an exoteric system of Saiva temple worship based on a well defined corpus of texts.11 Here Siva is worshipped as Sadasiva,12 and
6 8 9 10 11 12

See G Evans 1982 373 ff for a discussion of this and of proper names and problems of refer7 ence in general. See Sanderson 1985 200 ff with note 106, 1988/90 136-48 See Sanderson 1988/90 for an account of the divisions of the Saiva Tantras. For an outline of the Krama, see Sanderson 1988/90 151-2, 164-7 For the date of Abhinavagupta and other Saiva authors I rely on Sanderson, e.g. 1988/90 For a brief outline of the Saiva Siddhanta, see Sanderson 1988/90.136, 159-60 In the early Kashminan Siddhanta, Sadasiva is worshipped without a consort. Then ManonmanI appears as his consort on some occasions but not as part of the regular (nitya)

Praxis: Saiva Kashmir

59

Saiva Siddhanta ritual is in general more congruent with the Vedic world. To the Vedic world itself, from the point of view of the extreme Vedic ritualist, the Mimamsaka, Bhairava was simply the recipient (sarnpraddna) of a certain set of dubious rituals. To the other side of the Saiva middle ground, however, we meet with a cluster of charismatic cults based on the unorthodox Bhairavatantras. Here we find the gnostic non-dualism known as the Trika which Abhinavagupta and his followers tried to expound. Trika Saivism is briefly described as a system of ritual striving to assimilate the power of a triad (trika) of three goddesses: Para, Parapara, and Apara, the first mild and benevolent, the other two raging and terrifying. At the outset the Trika propagated the cult of these three goddesses alone, but this is not the only triad encompassed by the name. Indeed, it is the wide distribution of this model of triadic power which facilitates the ambition of Abhinavagupta's Trika to attain the state of a universal Saiva exegesis by the process of assimilating the older triads to the metaphysical conceptions which it gives to its own.13 This esoteric fringe set itself up as a separate division, offering what it considered a better way to salvation. It sought to incorporate the doctrines of the wilder Kaula14 and Krama practices, concerned in ritual with impurity, sex, and death, and centred around the goddess Kali and her emanations. At the turn of the millennium, then, Trika Saivism found itself competing with the dualistic and more conservative Saiva Siddhanta. Hence it tried to resurrect the wilder Bhairava within the cult of Svacchandabhairava, and to read a gnostic non-dualism into a basically dualistic corpus of texts. In other words, Abhinavagupta's task was to displace the dualistic exegesis of the Saiva Siddhanta, as well as to incorporate the more heterodox elements of the teaching he sought to propagate, while at the same time offering an exegesis that was acceptable within a general system of shared beliefs. Primarily Abhinavagupta's exegesis is a rather narrow exegesis of the MdlinTvijayottaratantra15 which he takes to be the fundamental scriptural authority of the Trika. He regards this as the summit and summation of all Saiva traditions, and develops a theory which accommodates the rest of Saivism within his own metaphysical system. Through his exegesis of the Mdlinivijayottara Abhinavagupta tries to cover a wide range of Saiva scripture and to estabpnvate ritual, normally only in the context of establishing external idols where the Sakti is the throne. ManonmanI is the highest of the nine Saktis that form his cosmic throne, and is placed m the centre (see H Brunner-Lachaux 1967 167-75, 1977 590 and note 42) In the later Siddhanta she appears as Sadasiva's regular consort 13 Sanderson (e g 1988/90 164) suggests that we may speak of at least three major phases m the evolution of the Trika The second phase sees the triad of the three goddesses subsumed within the goddess Kali The third phase Sanderson (ibid) describes as follows. 'Finally we have the Pratyabhyna-based Trika of Abhinavagupta with its two aspects, the first being the Kali-based cult of the Tantrdloka, and the second the condensed cult of Para as Solitary Heroine' (ekavird) See Sanderson 1988/90 for a detailed exposition 14 Developing from withm the Yogml cults, the erotico-mystical Kaula Saivism (Jcula, lit 'family; lineage' but later interpreted to mean 'body, totality') carried the Kapahka cult of power through impurity into the Krama and from there got incorporated into the Trika See Sanderson 1988/90 147-58 for an exposition of Kaula Saivism. 15 See, e.g , TA 117-18, and Sanderson 1992 292.

60

Indian semantic analysis

lish a basis of hermeneutics. His task in exegesis was first and foremost to incorporate into the non-dualist Trika the cult of the goddess Kali or KalasamkarsinI, Kali as the destroyer of time. The ritual and metaphysics of the MdlinTvijayottara, however, lack the Goddess as well as the idea that the universe is a projection in and of consciousness. Indeed, as A. Sanderson (1992) has pointed out, it also lacks the doctrine of non-dualism so fundamental to the Trika. Still, it is on this basis that Abhinavagupta seeks to establish that the various branches of Saiva literature embody modes of consciousness which collectively form a structure that perfectly reflects and therefore is the expression of the core of female powers which is inherent in Siva/Bhairava and therefore constitutes the essence of all phenomena. It is, however, important to remember that Abhinavagupta wrote against a ritual background. The Trika Saivism he propagates sought to direct its sect-defining ritual inwards in such a way that it could be abandoned altogether once the gnostic insight was no longer in need of ritual support. Indeed, this Saivism is ritual, and in its philosophical form ritual internalised. In its interiorisation it is thus strikingly similar to the way in which sacrifice was interiorised in late Vedic times.16 So, when Abhinavagupta's pupil Rajanaka Ksemaraja (fl.c. 1000-1050 CE) wrote his commentary on the Svacchandatantra Tantra of the Autonomous', he wrote it to defeat the prevalent adherence to the dualistic Saiva Siddhanta. The voluminous Svacchandatantra is in form a dialogue between Siva as Svacchandabhairava 'Autonomous Bhairava' and his Sakti Tower' as Bhairavl, and is mainly concerned with rituals of initiation and the desiderative practices of the Sadhaka. It is accordingly not particularly concerned with metaphysics, and no more non-dualistic than the MdlinTvijaya}1 In order to displace the dualistic exegesis of the text, however, Ksemaraja enters straight into its very name Svacchandatantra to demonstrate that its doctrine is nondualistic. At the end of his subsequently authoritative commentary, the Svacchandatantroddyota 'Illumination of the Tantra of the Autonomous', Ksemaraja writes (SvTU 6[15]: 146,11-16): ndmnaiva bhedadrstir vidhutd yendsvatantratdtattvd I srimatsvatantratantram bhedavydkhydm na tat sahate II bhedadarsanasamskdratantusantatam dditah I svacchasvacchandacitsvdtmasatattvam neksate janah II gatdnugatikaproktabhedavydkhydtamo 'panut I tenddvaitdmrtasphitah svacchandoddyota umbhitah II This illustrious Tantra of the Autonomous (Svatantratantra = Svacchandatantra) cannot tolerate dualistic exegesis since dualism, the nature of which is lack of autonomy (asvatantratd), is eliminated by its very name. People do not seeihat their identity is pure, autonomous consciousness, since it is enmeshed from the beginning by the threads of their deep conviction of the truth of dualism. Therefore this Svacchandoddyota (Illumination of the [Tantra of the] Autonomous) has been com16 17

A thorough analysis of meaning m Tantric ritual is provided by Sanderson 1995 See Sanderson 1986 208 and 1992.306

Praxis: Saiva Kashmir

61

posed, rich with the nectar of non-dualism, m order to dispel the darkness of the duahstic exegesis propagated by the blind followers of tradition. Ksemaraja thus leaves us in no doubt as to why he wrote his commentary on the Svacchandatantra. It is significant that according to him the very name of the text eliminates heteronomy, and so it cannot be subject to a dualistic interpretation. Ksemaraja makes a similar statement towards the end of his commentary on the Netratantra (NeTU 2[22]:343,21-2)18 on which too he imposed his non-dualistic exegesis. Enmeshed by the threads created by his dualist view, the Saiva Siddhantin is unable to see his true identity, to see the infinite autonomy of his own self, which ultimately is Bhairava, as revealed by Ksemaraja's non-dualist, idealist interpretation. With this background we shall see how these non-dualists were able to reveal the true nature of Bhairava by means of his name alone, so that the cult of an envigorated Bhairava could arise from its existence on the obscure boundaries of the Vedic world to its position within the high Brahmanical culture of Saiva Kashmir. A mangalasloka by Ksemaraja Ksemaraja, in the first mangalasloka - auspicious introductory verse - of his commentary on the Vijndnabhairavatantra,19 encapsulates the nature of Bhairava by referring in a highly condensed fashion to certain features of the deity's excellence. These features, as will be seen later, are not picked out at random, but rest on established explanations of Bhairava's name, explanations arrived at through the device of nirvacana analysis. It is impossible to appreciate the power of this verse as a statement in religious discourse, indeed to make sense of the verse at all, unless we are able to place it within its proper exegetical context. In order to amplify this, I shall first of all present and translate Ksemaraja's verse (VBhU 1,2-5): hhirundm abhayaprado bhavabhaydkrandasya hetus tato hrddhdmni prathitas ca bhiravarucdm Tso 'ntakasydntakah I bheram20 vdyati yah suyoginivahas21 tasya prabhur bhairavo visvasmin bharanddikrd vijayate vijhdnarupah parah221I
18 19 20

For the reading utthitah here, read umbhitah as m the SvTU passage above. Ksemaraja's commentary, the Vijndnabhairavoddyota, covers as we have it VBh 1-23. So read for bhiram The emendation is supported by the reading of the MahdrthamanjarTparimala which quotes Ksemaraja's verse (MMP 180,1-4), further by the nirvacana given at TA 1.98 and TAV 1(1)141,15, as well as by Sivopadhyaya at VBh 130 (KSTS 8:116,4). Last but not least it is supported by Ksemaraja's own commentary on the Svacchandatantra (SvTU 1[1]3,12 ff) The confusion of e and F is common among Kashmiri-speakers, see G.A Gnerson 1932 41,44-67 * Most Kashmiris seem to be unable to distinguish the two sounds' 21 So read for svayogf, cf MMP loc cit. The reading of sva° for su° is most certainly caused by the similarity in the Sarada script of -va and -u 22 MMP (loc cit) reads sivah for parah The two emendations suggested above, but not the present reading, are supported by an otherwise very corrupt transcription of this verse inserted after

62

Indian semantic analysis Bestowing absence of fear for those who are terrified, he is the cause of the crying out from fear of transmigratory existence, and from that [cry] he becomes manifest in the domain of the heart; he is the Lord of those who delight in their terrifying roaring, the death of death. Being the Master of that flock of excellent Yogins who blow on time, he is Bhairava; as the author of sustaining etc., he extends his Power throughout the universe, the Supreme whose form is consciousness.

What may at first sight emerge as a patchwork of incompatible elements makes perfect sense when viewed against the background of various nirvacanas current among the non-dualist Saivas of Kashmir. Indeed, the verse is a striking example of how the method of nirvacana analysis developed into an ingenious device in cultural discourse. All of the nirvacanas embedded in the verse are contained in Abhinavagupta's discussion at Tantrdloka 1.95-100. Some of them are even met with in the Vijndnabhairavatantra itself, or at least grounded on statements met with in that text. What the various nirvacanas have in common is that they all serve as valid introductions to the name 'Bhairava', thus bringing forth the very nature of the deity. This situation may serve to remind us that meaning is generated in the same way that language is generated, and thus subject to change and interpretation, that meaning, in fact, is created by people interpreting. This raises the following questions. What are the criteria for considering something an adequate introduction to a name? How do the explanatory expressions, whose semantic properties are a function of social practice, act in the minds of the individuals who use them? I have already argued that no belief can exist without being surrounded by a whole galaxy of beliefs, and that these beliefs fit together according to the rules of a certain logic in as much as people try to get as consistent a picture of the world as possible. The inner dependency of what words mean and how the world is ordered implies that changes in one may entail changes in the other. Now, it will be illustrated beyond doubt by the Tantrdloka passage dealing with the term 'Bhairava' that the nirvacana device enables those who mastered it to enforce or modify beliefs by encoding meaning into already existing terms. As pointed out earlier, meanings and beliefs are two sides of the same coin. There has to be a certain overlap in beliefs for people to talk about the same things, although these beliefs do not have to be identical. Still, would any nirvacana make an adequate introduction to a name simply by fact of its depending on some current belief? In simple Footnote 22 (cont) the final colophon of a Malayalam MS of the Isvarapratyabhijndkdrikdvrtti of Utpaladeva (Umv MSS Library, Tnvandrum, Ser. No 2317, Ace No. 8900 A (See Alphabetical Index of the Skt. MSS in the Univ. MSS Library, Tnvandrum, Vol 1 (A to NA), p 91)) sitikantham abhayaprado bhavadaydkandasya hetus tato bhutvdtiprathitas tadiya vacasdm Tso ntakasydntakah I bheram vdyati yah suyogmivahas tasya prabhur bhairavo visvasthidabharanddikrd vijayate vijhdnarupah parah// On this evidence I have retained the reading parah I owe this reference to Alexis Sanderson

Praxis: Saiva Kashmir

63

words: could a nirvacana be suggested if it were not resting on beliefs already common, and, conversely, if a nirvacana does rest on current ideas or beliefs, are there no limits to the applicability of such a device? These questions could be multiplied, but those posed already seem sufficient to demonstrate the necessity of investigating in detail the indigenous cultural underpinnings of the relation which holds between a given term and its corresponding explanatory expression. The Tantrdloka and Tantrdlokaviveka: anvartha revisited As a starting point I shall examine in detail the Tanlrdloka section dealing with the term 'Bhairava', together with its commentary Tantrdlokaviveka by Jayaratha,23 and then some passages of other texts of the Sanskrit Saiva literature which pertain to the analysis of 'Bhairava' as well as to the notion of nirvacana in general, passages which will be seen to shed some light on several of the questions involved. Jayaratha introduces the first verse of the Tantrdloka section in the following way (TAV 1[1] 139,7-8): bahusaktitvam eva ca etad abhidhdyakdndm pravrttinimittam ity aha . . . And it is this very possession of many Powers that is the ground for the use of the various [terms] which denote [Siva].24 Therefore he [= Abhmavagupta] states [in the next verse]: . . . Here Jayaratha tells us quite clearly that he considers the usage of terms which in their particular significations denote the deity to be grounded on the fact that He 24 possesses many Powers. In the preceding passage (TA 1.60c-94) Abhinavagupta has argued that Siva/Bhairava being absolutely autonomous is the sum and ground of all Powers, these, even in their most particularised form as individual phenomena or events in consciousness being no more than his single and essential power, his Power as Autonomous Agent (svdtantryasakti). That is to say, the various aspects which constitute his nature provide the grounds for using his various names. This implies, first of all, that what something means depends on what someone believes. A no less fundamental point is made by Abhinavagupta himself in his opening verse which reveals certain basic considerations completely on a par with the tradition of the Nirukta and nirvacanasdstra (TA 1.95):
23

Jayaratha (fl c. 1225-75 AD) was a representative of the Tnka-based Kaula cult of the goddess TnpurasundarT, or Srlvidya So was the much later Sivopadhyaya (fl c 1725-75) whose comments I shall examine below. For an outline of the cult of Tnpurasundarl, see Sanderson 1988/90 156-7 24 Alternatively, one could take etadabhidhayakanam to be a compound and translate 'and it is the very possession of many Powers that is the ground for the use of the various [terms] which denote Him' In this context etat 'this' would mean 'this [absolute]' which equals Paramasiva 'Supreme Siva'.

64

Indian semantic analysis devo hy anvarthasastroktaih sabdaih samupadisyate I mahdbhairavadevo 'yarn patir yah paramah sivah II

This is a statement of considerable importance. However, difficulties of interpretation to a certain extent vitiate a clear translation of the verse. As I read it devah in the first part is neutral or general ('the Deity'). The first half could thus tentatively be translated: Indeed [this nature of] the Deity is fully revealed {samupadisyate) through [the use of] terms which are inherently significant (anvartha) and traditional (sdstrokta)... I shall return to details concerning the translation of some of these words. The real difficulty is the interpretation of the word sabdaih which I have so far translated fairly neutrally 'through terms'. If this word is interpreted in the light of Jayaratha's abhidhdyakdndm, then the abhidhdyakdh sahddh 'the various terms which denote [Him]' are: (1) 'Bhairava', (2) 'Deva', (3) 'Pati', and (4) 'Siva'. This interpretation is supported by the fact that these are the terms contained in the latter half of Tantrdloka 1.95 as well as the fact that they are dealt with in the sections of the Tantrdloka immediately following the present verse. The anvarthasdstroktdh sabddh 'descriptive terms of the tradition' are thus [Maha-] Bhairava (TA 1.96-100), Deva (101-4b), p a t i (KHc-e 1 [up to tena patih]) and Siva (104e^ [sreyomaya eva sivo ndsivam kim api tatra]). The word sabdaih could, however, be taken to include .even the explanatory expressions, the nirvacanas, which are also descriptive and taught in sdstra. If we forget about the last mentioned difficulty for a moment and consider the verse as a syntactically coherent unit we may translate it in the following manner: Indeed [this nature of] the deity is fully exhibited/taught (samupadisyate) through [the use of] terms/names which are inherently significant (anvartha) and traditional (sdstrokta): this great 'Bhairava'-'Deva', who is the 'Pati', the Supreme 'Siva'. Although one may entertain a number of reservations about points of detail, the verse does in any case reveal highly interesting information. A clear interpretation depends, however, not only on the solution of the difficulties already mentioned, but also on the exact meaning of some of the words used, in the first place on the exact meaning of the verbal from samupadisyate. Turning to the major Sanskrit lexica, including the Vdcaspatyam, one finds that the verbal base sam-upaAdis is surprisingly rare. In fact, it is listed only in the dictionary of Monier-Williams (MW): 'P. -disati, to point out or indicate fully, show MBh.; to assign (with ace. of person and of thing), R.', the very few references being limited to the epics, and the meanings assigned not being particularly exact. Jayaratha does not remark on the word at all. To get closer to the meaning of sam-upaAdis it may thus be justified to start from the common upaAdis, with the assumption that the preverb sam carries the sense of samyak 'fully; thoroughly; rightly' or at least in some way the sense of reinforcement. Now,

Praxis: Saiva Kashmir

65

upa^dis moves semantically within the range of such meanings as 'point out; indicate; exhibit; teach; instruct'.25 Whatever type of terms or expressions the word sabdaih refers to, these are clearly said to be anvarthasdstrokta, which I take to be a dvandva. But of what kind is the relation sustained between the two words anvartha and sdstrokta] Is there intended to be an opposition here, so that the expressions referred to are either anvartha or sdstroktal The commentary of Jayaratha does not offer any clear answer to this at all, although his gloss is of considerable interest. His remarks on the entire verse run as follows (TAV 1[1] 139,13-14): anvarthaih vyutpannaih niruktaih sdstroktaih sdmayikaih II This agrees well with the interpretation as a dvandva compound. Abhinavagupta states that the deity is fully taught or referred to or fully exhibited by linguistic expressions that are anvartha, in accordance with that which they signify. Jayaratha, who glosses anvarthaih by vyutpannaih niruktaih, does not seem to distinguish sharply between the terms vyutpanna and nirukta. Strictly speaking, a vyutpatti is an analysis in accordance with that process of grammar by which a form is derived from radical elements by the addition of suffixes etc., while a nirvacana or nirukti goes into the tattva of the referent, the 'that-ness' or unchangeable essence of something by means of which it can be recognised, and so tells us why something is called what it is called. Words which are nirukta are thus capable of semantic analysis. A vyutpatti, on the other hand, does not necessarily help us to the real nature of that which is denoted by a word, it is just capable of producing the word in grammar with a basic semantic range thus provided. Accordingly, to say that a word is vyutpanna means that the word is capable of grammatical derivation. It may be that Jayaratha understands the term anvartha to include both vyutpanna and nirukta, because a regularly formed word like pdcaka 'a cook' is certainly anvartha and at the same time vyutpanna, analysable on a formal level. Not all terms that are nirukta are however vyutpanna. From the terminology of the present context, it seems thus reasonable to understand vyutpatti as the narrow term and nirvacana as the larger one. I consider, then, the terms vyutpannaih and niruktaih in Jayaratha's commentary to be related by a missing ca 'and' or vd samuccaye, '"or" in the conjunctive sense'. Otherwise one would have to understand niruktaih as a further gloss of vyutpannaih, a possibility one should certainly not rule out. Jayaratha's commentary continues: sdstroktaih sdmayikaih. This brings us back to the trickier bit of interpretation, namely the relation between the two * members of the compound anvarthasdstroktaih, that is, between anvartha and sdstrokta. Jayaratha seems to gloss sdstrokta 'taught in sdstra" by sdmayika which could be translated 'established by convention'. Now, is he contrasting
25

According to Panmlyas, the root ^dis with the preverb upa denotes the act of uttering, e g upadista 'uttered'; see Cardona 1988 640

66

Indian semantic analysis

sdstrokta with anvartha or is he not? I am inclined to think that he means to say that the terms in question are both anvartha and sdstrokta, in which case this is not a dvandva but a karmadhdraya.26 This seems to be right since these terms do not only happen to reflect the nature of that which they signify, so that 'Bhairava', as will be seen, actually tells us something about the nature of the deity. The fact that we use the term 'Bhairava' is not purely an arbitrary description, but it is something that is established by the' conventions of the sdstra - the verbal codification of the teaching in the tradition. On this basis anvarthaih and sdstroktaih ought to go together, and that is why we have no particles introduced here. Taking into consideration, then, that a nirvacana in grammatical literature is said to be anvartha, it is not unlikely that the statement sabdaih samupadisyate, 'is taught/fully exhibited through names/expressions/linguistic items', actually includes the various explanations of a term such as 'Bhairava', that is, the formulations of the nirvacanas themselves or the underlying elements which constitute the basis for these explanations, and that these terms too are sdstrokta in as much as they are both in accordance with the meaning and nature of that which they uniquely describe and are established through the verbal codification of the teaching in the tradition. But this creates an additional problem which begs for a solution. If it is the case that the form sabdaih refers to linguistic items in such a way that these include the explanatory nirvacana expressions, and that these are also to be sdstrokta, taught in sdstra, then the possibility for making up any new nirvacana of a term would seemingly be severely reduced. This would go against the very idea of nirvacana analysis. So it may be that what is referred to by the word sdstrokta is not only the strict formulation of a nirvacana in itself but also those ideas or predicates which are associated with a term in the tradition and more or less loosely hinted at by linguistic elements in scripture, those elements which constitute the basis for the full formal statement of^a nirvacana. Thus, while the representational properties of the term JBhairava' can be explained exclusively on the basis' of the term itself, this cannot happen independently of the ideas and purposes current among those who use the term. These ideas are sanctioned by the authorities of the tradition as is clear for example from Abhinavagupta's own statements27 that through nirvacanas a term is known by the teachers, that is to say, according to Jayaratha, by those who are responsible for the codification of various doctrines in the tradition. But these ideas may be challenged by changes in doctrine, so how?strong is this reliance on the authority of the tradition? Abhinavagupta on nirvacana analysis and reliance on tradition At Tantrdloka 6.30-3 Abhinavagupta makes it absolutely clear tKat he considers any term met with in scripture to be susceptible of nirvacana or nirukti
26

Cf the opemngwords of Yaska's Nirukta: samamnayah samamndtah, see p. 29, note 42, 27 above. TA 1.100 and 102; see below.

Praxis: Saiva Kashmir

67

analysis. At Tantrdloka 6.30 he offers two interpretations, dualist and nondualist respectively, of the term adhvan, literally 'route; way; channel'.28 The passage, together with Jayaratha's commentary on it, is worth discussing in full (TA 6.30/TAV 4[6]30,ll-31,8): nanv atrddhvasabdasya pravrttau kim nimittam ity dsankydha adhvd kramena ydtavye pade samprdptikdranam I dvaitindm bhogyabhdvdt tu prabuddhdndm yato 'dyatell 30// 'ydtavyepade' in sivatattvdtmani I bhedadasdydm hi tattattattvollanghanakramena sattnmsam sivatattvam prdpyatvenoktaml bhogyabhdvad ity adamyatvdt; adhigatasamvittattvd hi sarvam svdtmasdt kurvantiti bhdvahl tenddhvaivddhvd, adyata ity adhvd ceti II 30// What motivates the use of the word adhvanl To answer this he (= Abhinavagupta) says: The word adhvan denotes for the dualists the means of reaching a goal by stages But for the enlightened [i.e., the nonduahsts (prabuddhdndm)] an adhvan is so called because being the object of experience {bhogyabhdvdt) it is devoured (adyate). ydtavye pade 'the goal to be reached' means siva-tattva (Siva-Reality).29 Because m the duahstic system,30 the thirty-sixth [tattva], namely siva-tattva, is defined as that which is to be reached by means of an ascent through tattva after tattva. Because it is bhogya, that is, because it is to be devoured.31 The point is that those who have realised the reality of consciousness, turn everything into themselves (svdtmasdt kurvanti)?1 [So there are two nirvacanas:] adhvaiva adhvd, 'adhvan is just a path', and adyata ity adhvd, 'adhvan is so called because it is devoured'. So these are the two interpretations of adhvan: according to the first (dualist) adhvan simply occurs in its common sense of 'a path', according to the second (non-dualist) adhvan is interpreted by means of nirvacana analysis to mean that which is devoured, internalised. This context prompts Abhinavagupta to remark that one must always presuppose that scriptural terms are anvartha. The text continues (TA 6.31ab/TAV 4[6]31,9-32,2): nanu sarvasabddndm samayamdtrdd evdrthapratipddanam siddhyet, - iti kim atrdnu\nvartha]smaranena na hi sarvatraivaitat sambhaveft ity dsankydha iha sarvatra sabdondm anvartham carcayed yatah I yady apy arthdbhidhdne sabddndm trayi gatir yaugiki rudhd yogarudhd cal tatrdpi yaugikyd eva prddhdnyam sanimittam tatra tasydh pravrtteh, ata evdnyad dvayam In brief, the term denotes certain evolutionary units m the structure of Saivite cosmology On the concept of adhvan, see A Padoux 1963 261 ff, 1990 330 ff This form of Saivism operates with levels of reality (tattvas) where Siva is ultimately a reflection in and of a single, unified consciousness 30 The reading °dasaydm of bhedadasdydm is rather odd here One wonders whether it is not a corruption for darsane Either way the expression has to be taken as meaning 'dualism' 31 That is, because it is such that it cannot be internalised. 32 That is, they present to themselves everything, subjectify everything, internalise everything
29 28

68

Indian semantic analysis atraiva yathd katham cid antarbhdvaniyam yena sarvatraivdnvarthacdrcd pdram ydydtll Surely, the capacity of any word to convey its meaning may be established by mere convention. So what is the point of reminding us here that [a term] accords with that which it signifies, for this can certainly not be applicable in all cases? He (= Abhinavagupta) answers the objection by saying: Because in this system one must always analyse words in accordance with that which they signify. Although there are three ways in which words may denote their object, namely etymologically, by convention, and by combination of the two,33 nonetheless the etymological (yaugiki) is the predominant among them, because that one is caused.34 So it is that the other two (rudhd and yogarudhd) should be included within this, as far as it is possible by whatever means are available so that an analysis in accordance with [that which the term] signifies may be accomplished in all cases.

Abhinavagupta goes on to elaborate on this. He quotes the Trika scripture Nisdtanatantra35 on the three categories of words, and asserts that even though this particular nirvacana of the term adhvan does not occur in scripture, it is to be understood whenever the word is used, for, he argues, how much can be contained in any single book? The text runs as follows (TA 6.31cd-33/TAV 4[6]32,3-33,8): tad aha uktam snmannisdcdre samjndtra trividhd matdll 3111 naimittikiprasiddhd ca tathdnyd pdribhdsikil purvatve vd pradhdnam sydt tatrdntarbhdvayet tatah II32 // ato 'dhvasabdasyokteyam niruktir noditdpi cetl kva cit svabuddhyd sdpy Whya kiyal lekhyam hi pustake II 3311 prasiddheti, sanimittatve 'pi kva cid eva rudhehlyad uktam tatra 'samjnd hi trividhd jheyd sivasastresu sarvadal pdribhdsikanaimittT siddhd cdsau prasiddhibhdk 11 iha naimittiki samjhd nimittdt tu samdgatdT
33 35 l<

j '

As in pankaja 'lotus' - it 34 derived frompanka 'mud' andya '-born' but has a meaning estabis lished by convention. That is to say, its derivation is transparent. The text is known only from citations. Alexis Sanderson informs me that there is a surviving MS from the eleventh century in Nepal (NAK 1.1606) which has the title Nisisamcdra in the colophon. This is one of the names under which this text is cited (Nisdtaha, Nisdcdra, Naisamcdra, Nisisamcdra), but the work is not identical with the text as known from citations. The MS has similar subject matter to the Nisdtana {Naisamcdra) passages cited at TA 15.83c-97b for a particular view of the internal correspondence of external sites of worship, but the similar passage of the surviving MS disagrees with it in a number of important particulars in the list of sites and their internal location. In other words, the MS is obviously another redaction of the same sort of material. The text is probably to be classified as one of the Trika sources, and almost certainly a late Trika source in that it seems already to be introducing non-dualism.

Praxis: Saiva Kashmir

69

itil evam iha sarvasamjndndm nimittatdpy avasyam vdcyety atraivam uktam ity aha 'aid1 itil nanu yady evam tat sarvatraiva kasmdd evam noktam ity dsahkydha noditetyddil kiyad iti, na hy atra sabdavyutpddanam prastutam iti bhdvahll 33 // Therefore he (= Abhinavagupta) says: It has been stated in the Nisdcdra that for us a term is threefold: [it can be] occasioned36 (naimittiki), established [by convention] (prasiddhd), or technical (pdribhdsiki). 31 Furthermore,37 this is in descending order of importance.38 Therefore one should subsume under this [also the other two].39 32 So it is that I have stated this nirukti of the word adhvan. But even if I had not stated it, it might well have been inferred through your intelligence. For how much detail can one put into a book? 33 'Established', that is, by convention, in a particular, restricted sense, even though [the term is] occasioned.40 As has been said in that text, [i.e. in the Nisdtana]: For in the Saiva Tantras a term should always be understood as of three kinds: it is established as being respectively technical {pdribhdsiki), occasional (naimittiki) and conventional (prasiddhibhdk). Of these three an occasional term is one that arises from an occasion. So it has been said in this very text that all terms must thus necessarily be shown to be occasions41 [for their objects], so he says 'therefore'. If so, then how come he has not done this in the case of every word? To answer this objection he has the passage beginning 'if it were not given' (noditd). [On] 'how much' (kiyad iti) [etc.]: for here [in this work] it is not a treatise on language that has been undertaken, that is the point. This is a very important statement, reminding one of the characteristic way of arguing on the basis of what may have been contained in lost or inaccessible parts of the Veda.42 It was thus assumed in the tradition that the knowledge of humans is imperfect and subject to constant improvement. In the present context it would be dangerous from the exegete's point of view to
36

Or 'intentional', although 'occasioned' is really the strict sense. There is something in the occasion which occasions the word; there is something in an adhvan which causes the name, whereas there is nothing in John that causes the name 'John'. In this way causality enters the picture as well. 37 I am not quite happy about the vd because it does not fit in its proper sense of indicating an alternative. It could simply be that it is there to show connection with the preceding, that is to say that it be taken samuccaye, in the sense of 'and; moreover; furthermore'. 38 Lit.: 'in being former there is predominance' or 'to be earlier in this list is to be more important'; we really want here prddhdnyam 'predominance', so one has to take pradhdnam here in the sense of bhdva: in other words, purvatve pradhdnatvam. 39 That is, one should subsume under naimittikT as much as possible, one should try to include the other two, reduce the other two to this one. 40 As mpanka-ja 'lotus', lit. 'mud-born', derived from panka and ja thus still being sanimitta, having an occasion for its use in language, but also conventional. 41 It is tempting to emend nimittatd to sanimittatd since it is not the word which is a nimitta but its object. 42 Cf. Taittiriyabrdhmana 3.10.11.4: anantd vai veddh, 'infinite indeed are the Vedas'.

70

Indian semantic analysis

reverse this order of naimittikT, prasiddhd, and pdribhdsiki and say that the pdribhdsiki is the most important, because then you are closing the text to further analysis. The present passage thus provides us with Abhinavagupta's clear answers to some of the questions posed earlier. Of particular interest is his statement that one must always presuppose that scriptural terms are anvartha and thus, by implication, that any scriptural term is susceptible of nirvacana analysis. It is also evident from what Abhinavagupta says, that such analysis must not necessarily rest on scriptural evidence. It is at the same time obvious that Abhinavagupta does lean on the authority of the tradition when he can. Quite often he quotes or refers to earlier authors when presenting nirvacanas of various terms. The following example may serve to illustrate this. Immediately after his lengthy exposition of the name 'Bhairava', he goes on to analyse the word deva 'deity' (TA1.101-4). In this respect he relies first of all on various meanings of Vdiv met with in the Paninlya dhdtupdtha (Dhp 4.1): divu knddvijigisdvyavahdradyutistutimodamadasvapnakdntigatisu. The nirvacanas he gives do not correspond to every one of these senses. They do, however, correspond exactly to the senses ascribed to Vd/v by Abhinavagupta himself in his Pardtrimsikdvivarana (PTV 8,2): divu knddvijigTsdvyavahdradyutistutigatisu, 'the root divU occurs when the meanings "play", "urge to conquer (transcend)", "conduct", "radiance", "praise" or "movement" are to be denoted'. In the context this appears to be a quotation. Whether this indicates that the meaning entry of the dhdtupdtha employed by Abhinavagupta actually read like that, or that only these senses ( were of interest to him, is hard to tell. i Of greater significance is the fact that the section of nirvacanas of the terml 'Deva' 43 is a paraphrase, or possibly a quotation, of Brhaspati's Sivatanusdstra. Abhinavagupta states (TA 1.102): iti nirvacanaih sivatanusdstre guruhhih smrto devahl '! v

Thus, through [the aforementioned] nirvacanas, [the significance of the term] deva is recalled by the teacher44 in the Sivatanusdstra. Jayaratha (TAV 1[1] 146,11) identifies gurubhih 'by the teacher' as brhaspatipddaih 'by the revered Brhaspati'. This Brhaspati is the author of the
43

In the 'deciphered' form suggested by Jayaratha (TAV 1[1] 144-6) the nirvacanas of the word 'deva' are as follows* divyati [so read for devyati, cf the following] kridati (144,14). . divyati vijigisate (145,6) divyati vyavaharati (145,9) divyati dyotate dyotayati va (145,12) divyate stuyate (146,1) divyate jdndti prasarati ca iti vd devah (146,6). The latter nirvac ana draws on the 'hermeneutic rule' sarve gatyarthd jhdndrthdh which says that all words having their significance within the semantic field of 'movement' at the same time carry a sense within the field of 'knowledge' * Thus gati 'a going' is interpreted as jhdna and prasarana, and divyati gacchati may be rephrased as divyati jdndti prasarati ca, '[he is calle "deva" because he] divyati, that is, understands and pervades' 44 ddare bahuvacanam

Praxis: Saiva Kashmir

71

Sivatanusdstra, an early work of the Saiva Siddhanta in Kashmir. Along with Sadyojyoti, Brhaspati is referred to by Ramakantha45 as a Siddha to be venerated even by Gurus, but no manuscript has come to light in South India or Kashmir, and his work seems to be lost. It is, however, frequently quoted by Abhinavagupta, as in the present case where he relies on it for the analysis of the term 'deva\ One more instance of Abhinavagupta's reliance on the tradition deserves mention at this point. At Tantrdloka 5.135-6 he refers to the nirvacanas of 'Bhairava' met with in the Vijndnabhairavatantra 13046 by quoting the last part of that verse (TA 5.135d-6ab): bhairavo vydpako 'khile/l 135// iti bhairavasabdasya samtatoccdrandc chivah I One will become Siva by constantly contemplating in mental utterance the term 'Bhairava', understanding that it means: '... Bhairava is he who pervades the entire universe'. From this it is again evident that Abhinavagupta relies on the tradition and refers to nirvacanas found in scripture. Of even greater importance, however, is the claim here that one will become Siva by contemplating in mental utterance the term 'Bhairava' and the fact that this claim of the Vijndnabhairavatantra is given in the context of explaining Bhairava's name. It goes without saying, then, that the power attached to nirvacana analysis in the tradition is strong. In fact, it is exactly this reliance on the tradition whenever possible combined with the inherent capacity of nirvacana analysis to reveal the semantic content of any term which makes this a powerful tool in the hermeneutics of religious discourse. Bhairava restored So this is the basis on which Abhinavagupta puts his exegetical skills to use. With this background it is now possible to return to the Tantrdloka section that gives the various explanations of Bhairava's name and investigate in detail Abhinavagupta's exegetical procedure of reducing Bhairava into a set of Brahmanically acceptable metaphysical terms. As will become evident, the naked term 'Bhairava' is itself the sole medium Abhinavagupta needs to establish the very nature of the deity, since the deity is fully exhibited by expressions that are in accordance with its nature and codified by the teachings of the distinct but definitely interrelated Saiva traditions of Kashmir. Metaphysics are thus superficially not superimposed on his name, but discovered from within the name.
45 46

In the second mangalasloka of his Matangaparamesvardgamavrtti (ed N R Bhatt 1977 1) For further reference to this verse, cf pp 83f with notes 68-70 below

72

Indian semantic analysis

Deriving the substance of his arguments from fundamental tenets of Saiva doctrine, Abhinavagupta demonstrates by novel interpretations that the term 'Bhairava' may have been used to refer but without a fully acknowledged and proper understanding of the reference. Such a procedure becomes possible when there is a certain confusion in the name-using practice, so that in the Saiva community any statement that involves the name 'Bhairava* would be flawed even if the conversants should happen to be in full agreement. Abhinavagupta can take full advantage of this inherited confusion in usage. It enables him to establish a coherent practice by making true statements about the nature of the deity by means of its name - true because these statements rely at the same time on the verbal codification of the teaching in the tradition. In this way meaning is inter subjectively determined. Resting satisfied that he does not bring in anything outlandish, Abhinavagupta can even go beyond the referential feature and identify Bhairava as the absolute deity of Saivism in accordance with a general conception of the divine nature acceptable to the Saiva middle ground. From the platform he has secured in this way, he can proceed to the more heterodox elements of the teaching he seeks to propagate, in particular the incorporation of the cult of the goddess Kali/ Kalasamkarsinl. The first verse to tell us why the absolute deity is referred to by the term 'Bhairava', a wonderfully condensed piece of Sanskrit, begins as follows (TA 1.96): visvam bibharti posanadhdranayogena41 tena ca bhriyatel [He is called 'Bhairava' for the following reasons: because] he carries {bibharti) the universe, according to [his] nourishing [it] and supporting [it]; also [because] he is carried {bhriyate) by it. Jayaratha comments (TAV 1 [1] 141,1-4): bibharti - dhdrayati posayati ca svdtmabhittisamlagnatvena tadulldsandt I tena iti visvena, bhriyate iti dhdryate posyate ca — tasya visvamayatvenaiva sarvatra sphurandtl 'He carries', that is, he supports and nourishes [the universe], because he manifests it as fused with the surface of appearance which is his own existence. 'He is carried', that is, he is [also] sustained and nourished 'by that', that is, by the universe; because he is manifest in everything in as much as it is he that is embodied as the universe. The universe has no existence except as a reflection on the surface of Bhairava.48 So it is he that supports the universe; so it is he that gives the uni47

Ktdd posana0 for purana0. The commentary has posana. This could, of course, be a secondary gloss on purana, but the Dhdtupdtha leaves no doubt (Dhp 3.5): dubhrh dhdratiaposanyoh Sivopadhyaya, commenting upon Vijndnabhairavatantra 130 (VBhV 115,12), also reads posana0. 48 Cf., e g., TA 3 65 which concludes the argument that the universe is a reflection in consciousness:

Praxis: Saiva Kashmir

73

verse its life, nourishes it.49 But Bhairava can only be himself in as much as he is embodied in the universe, and so the universe sustains and nourishes his nature. To express this, Abhinavagupta draws on the well-established senses ascribed to the verbal root Abhr in the meaning entry of the Dhdtupdtha. By interpreting 'Bhairava' as both agent and object in relation to the activity denoted by ^Ibhr, that is, as kartrsddhana as well as karmasddhana, he brings out the essential doctrine of Trika Saivism that the absolute is both transcendent and immanent. This nirvacana is rendered complete by the continuation of the verse (TA 1.96): savimarsatayd ravarupatas ca Moreover, because his form is rava, the roar, by fact of his being endowed with selfcognition. Bhairava has awareness or ideation, vimarsa, as his dynamic essence. And that ideation in its most subtle form is AHAM T , a self-awareness which is the supreme mantravirya, the inner essence of mantras,50 or rava 'the roar'. What we have here, then, is a nirvacana on the basis of two verbal activities expressed by two verbal roots, ^Ibhr 'nourish; support' and Vrw 'sound'.51 Jayaratha explains this as follows in his commentary (TAV 1[1] 141,4-5): ravarupatah iti sabdanasvdbhdvyat, tena bharandd ravandc ca bhairavah ity ayam niruktahl 'Because his form is rava (roar)', that is, because his nature is [the activity of] sounding. So this [term] 'Bhairava' has been semantically analysed (niruktah) on the basis of [the activities of] supporting/nourishing and roaring. The word sabdanam would most easily lend itself to an interpretation as bhdvasddhana, that is, as a noun denoting the mere activity of sounding. This would imply that Bhairava is the sounding. It is also possible to interpret Footnote 48 (cont.) ittham visvam idam ndthe bhairaviyacidambare I pratibimbam alam svacche na khalv anyaprasddatah II Thus this universe is a reflection in the perfectly transparent void of the consciousness of Bhairava, the Lord. And it exists as a reflection without depending upon anything else. 49 Cf. BPD 4. sa eva bhairavo devojagadbharanalaksanah/ svdtmddarse samagram hi yacchaktyd pratibimbitam II This [controller of Power (saktimari)] is the deity Bhairava, [so called because] he is defined by sustaining (bharana) the universe, for by his Power he has caused everything to appear in that mirror which is his own identity. 50 On the concept of the mantravirya, see A. Padoux 1963:303 ff.; 1990:383 ff. 51 Dhp 2.24* ru sabde. Incidentally, the name 'Rudra' is analysed by the same verbal root at Nirukta 10.5; cf. Kahrs 1980:248.

74

Indian semantic analysis

sabdanasvdbhdvydt as 'because it is his nature to produce sound'. The distinction is however rendered less crucial since there is no difference between agent and object (kartrkarmdbheda) with regard to Bhairava here, moreover because sabdanasvdbhdvydt in this special context means that Bhairava is aware of himself in this inner language which is the instinct of consciousness, the instinct of the light of reality. So his nature is a constant roaring of the great mantra of T , AHAM.52 Thus the term 'Bhairava' refers to unconditioned subjectivity as the essence of all phenomena. A similar analysis is met with also in the Paratrimsikdvivarana where Abhinavagupta presents it in the following form (PTV 63,11-12): bhairavo bharandtmako mahdmantraravdtmakas ca, 'Bhairava is he whose nature is carrying (bharanam) and the roar (rava°) which is the great mantra (AHAM)\ So he is called 'Bhairava' because he supports the universe and is supported by it, because he nourishes it and is nourished by it (bharandt: ^bhr), and because he roars (ravandt: Vrw), being the tanscendental subject that internalises the universe. Now follows another analysis (TA 1.96): samsdrabhiruhitakrc call 96// [He is] also [called Bhairava because] he favours those terrified (bhlru) by samsdra. Jayaratha elaborates on this analysis as follows (TAV 1[1] 141,5-7): bhirundm ayam hitakrt iti bhairdvah, bhirutve ca nimittam samsdrah tena samsdnndm abhayapradah ity arthahl He is one who favours those who are terrified: thus [he is called] Bhairava. And the cause of this state of terror is samsdra, the cycle of births and deaths. That is to say, therefore, that he bestows absence of fear upon those who are trapped in samsdra. Bhairava is he who frees from fear those trapped in samsdra, the endless cycle of births and deaths. The basic element here is bhiru, on^which the name 'bhairava' is formed as a secondary derivation. Now, this is also the acceptable analysis according to a Paninlya grammatical derivation: A 3.2.174 bhiyah kruklukanau teaches that the krt suffixes (A 3.1.1 pratyayah) Kru (-ru) and KlukaN (-luka with presuffixal accent) are added after (A 3.1.2 paras ca) the verbal root (A 3.1.91 dhdtoh) bhl 'fear' to form primary derivations denoting an agent (A 3.4.67 kartari krt) on the condition that he performs the activity of fearing as part of his nature, either habitually or as a duty, or that he performs an act well (A 3.2.134 a kves tacchilataddharmatatsddhukdrisu). This gives us the krt or primary formation bhiru in the sense 'whose habit or duty or skill it is to be afraid'. From this we get the taddhita or secondary formation bhairava according to the following derivation:
52

Cf TA 3 221c-3b Summarising his exposition of the matrkdeakra (from a to ha with the bindu [am] containing all the phonemes), Abhinavagupta explains how AHAM (A-HA-M) denotes/is the totality. Here we border on ma^ra-analysis.

Praxis: Saiva Kashmir A 4.3.120 tasyedam i A 7.2.117 taddhitesv acdm ddeh 4 ' A 6.4.146 or gunah i ' A 6.1.78 eco 'yavdydvah bhiru aN i i ai >t 4 o I av bhairava

75

A 4.1.83 prdg divyato 'n teaches that the taddhita suffix aN {-a with vrddhi strengthening of the first vowel) is added under meaning conditions given in rules up to rule A 4.4.2 tena divyati khanatijayatijitam. That is to say, A 4.1.83 is a general rule (utsarga) which teaches the addition of a/*/unless it is blocked by some other suffix under conditions specified by a special rule (apavdda). Rule A 4.3.120 tasyedam teaches the addition of aN in the sense 'this pertains to it'. The taddhita suffix aN is thus very loose in sense and covers almost any relation (sambandhamdtra) in as much as it denotes whatever relation can be marked by a genitive case suffix. It is worthy of notice that Abhinavagupta exploits this fact that the suffix aN is very general in sense, and he can thus specify its meaning to suit his own purposes: samsdrabhiruhitakrt 'he favours those whose habit it is to be afraid because of samsdra'.53 But surely a regular grammatical analysis of 'bhairava' does not prevent further analysis of the name which is quite rich in possibilities. Here again nirukta adds to the possible meanings arrived at by vydkarana.5A Still another sense of the term is revealed by the next verse based on the fact that pardmarsa 'self-awareness' can be taken to be the sense of rava because the former is the sound of self-awareness (see, e.g., TAV 3[4]205,5-6) (TA 1.97): samsdrabhitijanitdd ravdt pardmarsato 'pi hrdijdtahl [He is referred to as Bhairava because] he is born in the heart from the roar or55 awareness produced by terror of transmigratory existence. The terror of what it is to be trapped in this endless cycle of births and deaths produces a roar or subjective awareness which makes Bhairava arise in the heart. So the analysis is in this case: bhT(ti)-rava-jdtah, that is, Bhairava is he
53

The wording °hitakrt might suggest that Abhinavagupta had in mind A 5 1 5 tasmai hitam which teaches taddhita formations in the sense 'good or beneficial for it', implying that Bhairava is he who is good for the afraid This would account for hitakrt as his chosen wording The problem is of course that bhiriindm is a genitive, not a dative form. In any case, the analysis here involves a derivation with bhiru and a taddhita suffix -a that triggers vrddhi strengthening of the first syllable This alternative will not work, though, because aN does not extend as far as A 5 1 5, that is to say, it is not taught there 54 Cf Yaska's statement at Nirukta 115 where he characterises nirvacanasastra as vydkaranasya kdrtsnyam svdrthasddhakam ca, 'a complement to grammar, moreover something which is a means to its own end', see p. 32 above. 55 The particle api in the sense of vd 'or', according to Jayaratha.

76

Indian semantic analysis

who is born from a roar of fear. Jayaratha comments on this as follows (TAV bhih56 samsdratrdsah, tayd janito yo 'sau ravah bhagavadvisaya dkrandah pardmarso vd tato jdtah iti bhairavah, tendkrandavatdm pardmarsavatdm ca hrdi paramdrthabhumau sphuritah iti ydvatl [The word] bhi means fear, that is, terror of transmigration. The roar which is produced by that [fear] means a crying out to the Lord, or rather, awareness [of the Lord]. [And] being born from that [cry or that awareness, he is called] Bhairava. So this means that [Bhairava] is manifest within the heart - that is, on the plane of ultimate reality - of those who call out in terror or those who direct their awareness [to the Lord]. Becoming aware of the fear of* what it is to be trapped in samsdra, that endless cycle of births and deaths, produces a crying out to the Lord, or rather, it produces an awareness of the Lord, an awareness which turns to the Lord as a result of this terror. Bhairava is thus conceived of as the object of awareness and as produced in the heart as a result of that awareness. In terms of kdraka analysis, then, we have in this case an interpretation as karmasddhana, that is, as the object of the activity expressed by the verbal root associated with the name 'Bhairava'. Abhinavagupta offers further explanations of the term in the continuation of the verse (TA 1.97): / prakatibhutam bhavabhayavimarsanam saktipdtato yena II 97 // / ['Bhairava' means] he by whom awareness of the fear of finite, transmigratory exis-i tence (bhava) is made manifest through the descent of Power. , Jayaratha gives the following elucidation (TAV 1 [1] 141,11—14): bhavdd bhayam bhis tasya ravo vivecanam vimarsanam tasya saktipdtamukhena ayam kdranam iti bhairavah, samsdravaimukhye 'pi ayam eva nimittam iti bhdvahl He is the cause - through the descent of Power - of that roar {ravah), that discrimination, that awareness of bhayam - that is, fear - of transmigratory existence; hence [he is called] Bhairava. The sense is that he alone is instrumental even in one's turning away from samsdra. The underlying analysis is in this case: bhiravasya kdranam iti bhairavah, '[he is] the cause of the fear-cry, hence "Bhairava"'. Bhairava is instrumental in producing the roar of fear, or rather, the discrimination or understanding of that terror which arises from the state of transmigration and causes - through
56

The text reads bhayam bhih. If bhayam is a gloss on bhih it is both superfluous and in the wrong place. Jayaratha has lifted this from Ksemaraja's Uddyota at the beginning of the Svacchandatantra (SvTU 1[1]:3,14), but there as well it is likely to be a corruption of the last word of the preceding clause. What we have above is then an example of an insertion of an interlinear gloss. For a discussion of the Ksemaraja passage, see p. 87 with note 74 below.

Praxis: Saiva Kashmir

77

the descent of Power, that is, grace - one's reversion from the state of selfrepresentation as a transmigratory being. Now we have a very elegant nirvacana coming up (TA 1.98): naksatraprerakakdlatattvasamsosakdrino ye cal kdlagrdsasamddhdnarasikamanahsu tesu caprakatah/l 98// ['Bhairava' means] he who is manifest among those whose awareness is devoted to that trance which devours time, that is, in those who desiccate that essence of time which is the propellor of the celestial bodies. Here the term has been divided up as follows: ([(bha \=naksatra\ + lra — > bhera) + va] — bherava) — bhairava, thus derived as a vrddhi formation of > > bherava. The item bha, which is taken to mean naksatra 'celestial body/bodies; constellation; asterism', is combined with the item Ira in the sense of preraka 'motivator; propellor' (Vzr 'move; rise; agitate') to form the item bhera and hence a synonym of 'time'. From bhera we get bherava by adding the item va, obviously thought of as related to the verbal root va 'blow; move'.57 For these bheravdh are further specified as samsosakdrinah, the desiccators of bhera 'time': as they blow on it they dry it out completely. They are the yogins who desiccate time by having their attention centred in that trance which is called the devouring of time. The master among them is Bhairava, the supreme Bherava, the supreme desiccator of time, being vividly manifest as that which devours time, bhairava thus derived by vowel-strengthening {vrddhi) of bherava. The linguistic items which provide the basis for this elaborate analysis of the term 'Bhairava' are hardly mentioned in the verse itself but are identified by Jayaratha (TAV 1[1] 141,14-142,4): bhdni naksatrdni irayaii prerayati iti bherah kdlah tasya tattvam ksanddydtmakam svarupam, tasya samyan nihsesena sosam abhibhavam kurvanti iti kdlam vdyanti iti bheravdh - kdlagrdsasamddhirudhdvadhdnd yoginah, tesu ayam svdmltattvena prakatah sphuritah iti bhairavah I It causes to move (irayati) - or propels {prerayati) - the bhdni, that is, the heavenly bodies, hence [it is called] bhera, time. Those who blow on time's essence {kdlam [=bheram] vdyanti), that is, on its nature consisting of moments etc., that is, those who completely {sam = samyak=nihsesena) dry it out {sosam kurvanti = ,* sosakdrinah) in the sense that they conquer {abhibhavam kurvanti) it, those are the bheravdh, those Yogins whose attention is centered in that samddhi which is called kdlagrdsa, the devouring of time. So that master among them, who is ^ Dhp 2.41: vd gatigandhanayoh, 'the root vd occurs when the meanings "moving" and "blowing (spreading of fragrance)" are to be denoted'; cf. also Dhp 1.969 ovai sosane, 'the root vai - occurs when the meaning "drying up" is to be denoted': the wind dries something while s blowing. Historically, the two roots were at one stage one and the same. Presumably, the shortening of vd to va above involves A 3.2.3 ato 'nupasarge kah which teaches that the krt suffix Ka (-a) is added after a verbal root which ends in a when it is not preceded by a preverb but C v co-occurs with a noun that denotes its direct object This accounts for forms such as goda'giverof cow(s)'.

78

Indian semantic analysis vividly manifest - that is, who has burst into view - as that [which devours time (tattvena)], [is accordingly called] Bhairava.

To sum up, bhera 'time' is that which propels the heavenly bodies. And those Yogins who desiccate time by having their attention centred in the samddhi called 'the devouring of time' are called the bheravdh. The master among them is Bhairava. He is the supreme Bherava, the supreme desiccator of time, being vividly manifest as that which devours it. While the preceding explanations represent fairly general Trika doctrine, the following analysis leads into the Krama system. As mentioned already, the term Krama Saivism refers to a cluster of extreme, charismatic cults centred around the goddess Kali and her emanations. It is that part of Kaula Saivism which has most clearly retained Kapalika elements. At the time of Abhinavagupta, Krama doctrine was permeating the Trika. In fact, the incorporation of Krama elements into the Trika is partly what Abhinavagupta wanted to achieve. The strong presence of the female in the esoteric liturgies of the Krama is that feature which most clearly disagrees with the Saiva Siddhanta where Siva was, at least originally, worshipped without his consort, that is, without any personification of his female Power (sakti). So, in the next explanation Abhinavagupta introduces the Krarna doctrine of the goddesses Khecarl etc. (TA 1.99-100): samkocipasujanabhiye ydsdm ravanam svakaranadevindm I antarbahiscaturvidhakhecaryddikaganasydpi II99 // tasya svdmT samsdravrttivighatanamahdbhimah I

,

It is difficult to make sense of this verse without including the first half of the subsequent verse, or at least its opening words tasya svdmT 'the master of that'. Stressing the item -bhi- in the word mahdbhimah 'most terrible', one might possibly understand the compound samsdravrttivighatanamahdbhimah to represent a separate analysis, although I consider this to be an unlikely possibility. I shall return to this point below. As it stands, one may ihterpret the text in the following way: ['Bhairava' means] he who, being most terrible in the act of shattering those processes which constitute the cycle of births and deaths, is the master of those goddesses of one's faculties whose roar causes terror to contracted, bound individuals, and [who is the master] also of the fourfold collection within and without of [the goddesses called] Khecarl etc. These goddesses are essential in the Krama deification of the cycle of cognition. So we see it in the commentary of Jayaratha (TAV 1[1] 142,4-10): samkocino bhedaprathdmayasya pasujanasya bhiye tattatsukhaduhkhddyupajananatrdsdya ravanam sabdardsisamutthakddikaldvimarsamayo ravo ydsdm tdh svakaranadevya indriyasaktayah, tathd antarbahih pramdtrprameyddydtmd caturvidhah catusprakdrah khecaryddiko ganah khecari-gocan-dikcan-bhucaryo bhiravds tdsdm ayam svdmlbhairavah I

Praxis: Saiva Kashmir

79

Those whose roaring (ydsdm ravanam) - that is, that roar which is their awareness of these various elements such as [the sound] ka arising from the totality of sounds - causes fear (bhiye) - namely that terror which gives rise to the various emotional states of pleasure, pain, etc. - to those who are in the state of contraction (samkocinah), that is, embodying the manifestation of plurality, to bound individuals (pasujanasya) - [namely] these goddesses of one's faculties (svakaranadevyah) - that is, Powers of the senses - and likewise that internal and external (antarbahih) fourfold, that is, quadruply-moded group beginning with Khecarl, that is to say, Khecarl, Gocarl, Dikcarl and Bhucarl, embodying [as it does] the agent of cognition, the object of cognition, etc. - those are the Bhlravas {bhiravdh). [And] this master (svdmi) of theirs is Bhairava. This requires some explanation. Bhairava is said here to be the master of the bhiravdh. Hence he is termed bhairavah 'the master of those whose roar causes fear'. These Bhlravas are firstly the goddesses of one's faculties. Now, why are they called Bhlravas? Because their active state in consciousness, that is, that roar (rava) which is their awareness of the energies that arise from the various sounds of the language, causes fear (bhi) to bound souls who are in the state of contraction, that is, to those who embody the manifestation of plurality. And what is this fear (bht)l It is, says Jayaratha, that terror which gives rise to the various states of pleasure, pain, etc. The idea here is that the sense-faculties, or rather the faculties in general the faculties of action, cognition, and internal cognition - express themselves in consciousness through the sounds of the language. The sounds of the language exist in a subtle state. As such they have been expounded at length by Abhinavagupta.58 But their gross combination gives rise in consciousness to the various emotional states. Thus there are the ideations or representations embodied in the series of letters, as opposed to the represented or external, objective world. The distinction between vdcya 'expressed' and vdcaka 'expressor'59 corresponds in the metaphysics of the Trika to the distinction between the modes of prakdsa 'manifestation; light' which is Siva and vimarsa 'awareness; ideation', the dynamic essence or Sakti. So it is the sounding forth of those Powers of one's faculties in consciousness which causes terror. And that is why they are bhi-ravdh - 'they whose roar causes terror'.60 So that is the first group of Powers that is called bhiravdh, those whose roar causes terror. The second group is the fourfold collection beginning with khecari, that is to say, the Powers of Khecarl, Gocarl, Dikcarl and Bhucarl.
58 59

TA 3 66 ff. Cf. Jayaratha TAV 2[3]74,12: iha hi visvasya vdcyavdcakdtmand dvidhd avabhdsah, 'for in t system the universe is represented in two modes, as expressed and as expressor'. 60 Cf. SivasutravimarsinT on Sivasutra 1.3-4 (SSV 16,11-17,2) Ksemaraja explains how the mdtrkd, an order of the sounds of the Sanskrit language from a to ksa and considered to represent Siva, gives rise to these various limited forms of knowledge such as 'I am incomplete', 'I am thin', 'I am fat', 'I am one who has performed the Agmstoma sacrifice'. Through permeating the various terms which denote those states, they assume the form of grief, surprise, joy, attachment, etc.

80

Indian semantic analysis

These are said in Abhinavagupta's verse to be both internal and external, which Jayaratha explains as 'embodying the agent of cognition, the object of cognition, etc ' What he is getting at is the following Khecarl (She who pervades the void61) is the Power of consciousness in the sphere of the parah pramdtd, the unlimited agent of cognition Gocarl (She who pervades speech62) is the energy of consciousness in the internal faculties Dikcarl (She who pervades the [ten] directions) is the Power of consciousness in the faculties of cognition and action, the ten indnyas Bhucarl (She who pervades the field) is the same Power of consciousness embodied in the external objects of perception So the sequence of the four sets of energies - Khecarl, Gocarl, Dikcarl &nd Bhucarl - represents the expansion of consciousness m the process of cognition from the inner state of the subject, through the internal faculties of the mind, into the external faculties of cognition and action, and finally into the objects themselves In the Krama system these four Saktis are transcended by a fifth called Vyomavamesvarl (the Lady who emits the void[s]), who represents the absolutely transcendent state before any projection of agency of cognition 63 These Saktis are highly developed in the Spanda-texts which have a very direct link with the Kali cult Being a kind of almost sect-neutral exegesis of the Kali tradition, these texts have a very strong connection with the Krama 64 With Vyomavamesvarl as the first, Khecarl and Bhucarl are the second and third among five goddesses which in Krama ritual represent the cycle of cogni61 62

That is bodhagagana 'the sky of consciousness', according to the Spandamrnaya (SN 38,5) Literally 'cows' In the Nighantu (111) the word gauh is listed among vannamdm, the names for speech Ksemaraja explains in the Spandamrnaya (SN 38,9) gaur vak tadupalaksita.su samjalpamayisu buddhyahamkaramanobhumisu carantyo gocaryah, 'the goqaryah called because] they are pervading/moving within the levels of the three faculties of internal cognition - the mtrovertive and extrovertive mental organs and the ego awareness, [levels] of conversation implied by gauh that is speech', cf Vairagarbha's commentary on the Hevajratantra quoted and translated as follows by D Snellgrove (1959, ptl 86, note 2) Likewise by saying that these are the five faculties of sense the eye and so on, with the name of cow (go), &c ' 63 This Sakti is also known as Vamesvari, cf Ksemaraja in his Pratyabhijnahrdaya (PH 28,12-29,4) kim ca citisaktir eva bhagavati visvavamanat samsaravamacaratvac ca vamesvarya sati khecan gocan dikcan bhucanrupaih asesaih pramatr antahkarana bahiskarana bhavasvabhavaih pansphurantT0 Moreover, this divine Power of consciousness being called Vamesvari - because of its pouring forth (vamanat) the umverse and because of its moving against the current (vamacaratvat) of [the extnnsicist perception which constitutes] samsara - radiates as all agents of cognition, inner faculties, external faculties, and external objects, [these radiations in these four spheres are] in the form of Khecarl, Gocan Dikcarl and Bhucarl [respectively] The text goes on to explain how this radiation takes place 64 See, for example, the Spandamrnaya of Ksemaraja where these Powers of Khecarl, Gocarl, Dikcarl and Bhucarl are dealt with (SN 38,21) There they are mentioned m the plural (khe caryah, etc ) and collectively referred to as four wheels of divinities, catvan devatdcakrani (ibid 38,3) Abhmavagupta himself may be cited, however He deals with these Powers in the Paratnmsikdvivarana (PTV 39,12-40,3) and m the Paryantapahcasikd 26-8 (see K C Pandey 1963 84)

Praxis Saiva Kashmir

81

tion from initial to final voidness, according to the Mahdnayaprakdsa (MNP 64,4-8) 65 So those too are the bhlravdh, those whose roar causes terror Now, the idea is that these same Powers can be either binding or liberating 66 Consciousness is always the embodiment of these four levels, but those energies can bind you or release you They will bind you if you do not realise that all appearance is the dynamics of consciousness, they will release you once this realisation is effected And so, in the worldly state, the roar of the Bhlravas does indeed lead to terror, to the terror of reincarnation, to the terror of limited states of consciousness * Bhairava is the master of these Bhlravas, and it is for this very reason that he is referred to by the term 'Bhairava' He is the material cause and at the same time the efficient cause of this deployment of Powers Thus, when you are in control of those Powers, when you realise that they are simply the •shining of consciousness, you are no longer contaminated by them you are Bhairava, since that is the state of Bhairava You are then the Cakresvara, the 'Lord of the Wheel of these Powers And in this way Bhairava is considered IsamsdravrttivighatanamahdbhTmah, 'he who is most terrible m the act of Shattering those processes which constitute the cycle of births and deaths' So 1 take this expression to give the reason for Bhairava's being the master of ithese two groups of goddesses or Powers whose roaring causes fear He is the jnost terrible, mahdbhimah, among them Jayaratha has very little to say here (TAV 1 [1] 142,10-11) mahdbhima iti bhisanah, tendtra bhairavasabdah samketitah iti bhavahll [The word] mahdbhimah [simply means] terrifying, [and] so it is that the term 'Bhairava' is agreed upon here, that is the point These remarks do not suggest that the expression mahdbhimah represents an explanation in its own right, that is, that it should be considered an explana%on separate from the preceding words tasya svdmT and the connected verse J)9 Other factors too warrant the conclusion that this was not Abhmavagupta's intention | First of all, the word mahdBHImah would by itself suggest an explanation $f the element bhai- only, not of the element -rava One might supply ^AVArupatah (from verse 96) or some similar expression that would account | 5 See also the Mahdrthamahjanpanmala of Mahesvarananda (MMP 89,4-90,6) Khecari and Bhucarl are also met with in esoteric Buddhism for example in the Hevajratantra At the zenith and nadir with regard to the wheel of yogmls presented there, they represent samsara and nirvana (Snellgrove 1959, pt 1 74), and are associated with the spheres of thought and touch respectively (ibid 126-7) One would thus have expected Khecari to represent nirvana and Bhucarl to represent samsara This may be parallel to Khecari being the Power of consciousness in the sphere of parah pramata, the unlimited agent of cognition, while Bhucarl is the same Power in the sphere of external objects According to Snellgrove (ibid 28,128), however, the 'formula of personality' associated with the Yoginis Khecari and Bhucarl in the Hevajratantra is speech and body respectively ?6 Cf the Spandamrnaya of Ksemaraja (SN 38,17-21)

82

Indian semantic analysis

for the latter half of the term 'Bhairava'. But that would not be in accordance with Abhinavagupta's ordinary practice, a nirvacana usually being stated in full, or, if not, being easily inferred from what has actually bee;i enunciated. Second, the expression mahdbhimah does not add any knowledge not already revealed by the previous explanations. Third, there is the fact that Jayaratha's preceding words tdsdm ayam svdmT bhairavah refer to that which is signified by the term 'Bhairava', not to the term itself. Jayaratha's usual procedure in concluding his elucidation of each nirvacana is to state Hti bhairavaK. Fourth, the text here can easily be read as a continuous whole, the editor's danda being superfluous:'... this master of theirs is Bhairava, [because he is] mahdbhimah, that is to say, terrifying, [and] therefore the term "Bhairava" is agreed upon here'. The latter part of his remark does explicitly refer to the term itself. It is the fact that he is mahdbhimah 'most terrible' which renders possible the previous explanation of the term 'Bhairava': 'so it is that the term "Bhairava" is agreed upon here; that is the point [of stating °mahdbhimahy. Incidentally, one might ask whether samketitah 'agreed upon; used according to convention (samketaY here is referring us back to the previous sdstroktaih sdmayikaih67 thus indicating that the word is used in a slightly more limited sense. This, however, is difficult to ascertain. What has emerged from the above analysis of the Tantrdloka passage is as many as six different nirvacanas of the term 'Bhairava'. These may conveniently be summed up by the following six structures, indicating the fundamental elements of each explanation. The linguistic items referring to the activities or relations considered basic to each specific analysis are pointed out within brackets: 1. 2. 3. 4. 5. bharandd ravandc ca: (BHARANA: ^BHR+RAVANA: bhirundm hitakrt: (BHIRU+the taddhita suffix aN). hhijanitdd ravdj jdtah: (BHI + RAVA + the taddhita suffix oAf). bhiravasya kdranam: (BHI+RAVA + the taddhita suffix aN). bhdni naksatrdmrayatiti bherah kdlas tarn vdyanti (sosayanti) iti bheravdh (kdlagrdsarasikd ye yoginas) tesdm svdmT: (((BHA + IRA) BHERA + VA) BHERAVA + the taddhita suffix 'aN). 6. bhiye ravanam ydsdm td bhfravd indriyadevyah khecaryddisaktayas ca tdsdm svdmT mahdbhimah: ((BHI + RAVA) BHIRAVA + the taddhita suffix aN).

Since the term 'Bhairava' refers to the absolute deity, it is to a Certain extent an ill-chosen example with regard to illustrating ^rate-analysis, that is, the assignment of categories according to how the referent is thought of as a participant in the activities associated with the term. In this respect the alternatives indicated by each analysis ate not always entirely clear. While the term 'Bhairava' refers to unconditioned subjectivity as the cause and substance
67

TA 1 95; TAV 1[1] 139,13-14, cf p. 65 above.

l

'

Praxis: Saiva Kashmir

83

of all phenomena, we nevertheless meet with the persistence of a theistic code in metaphysics which evinces a conception of the deity as an ideal selfrepresentation. It is thus not surprising that various interpretations of 'Bhairava' as a noun denoting an agent (kartf) or object (karman) are suggested rather than an interpretation as an action noun denoting mere being or state of action (bhava). The different fcarafca-interpretations seek to establish a reality which is to be recognised through contemplation by the devotee. And so the term 'Bhairava' is subject to fcarafaz-analysis just like any other noun to the extent there exist activities argued to be the causes for the use of the term. Thus the elements bhT 'fear' and rava 'roar' are taken as the basis for interpretations of 'Bhairava' as, for example, karman (3) - he is produced from a roar of fear, karana 'instrument7 (4) - he is the means of producing that roar of fear which turns one away from transmigratory existence, or kartr (5) - he is the supreme Bherava, the supreme desiccator of time. Through exegesis of his name alone, Bhairava is thus established as the universal object of awareness and worship, while the ultimate identity of the individual is reality itself as an unconditioned Power of self-representation which is ' nothing but Bhairava, the Self who is the identity of Selves. Abhinavagupta states in his Bodhapancadasikd (BPD lied): sa eva khalu samsdro yo vibhisakah, 'the cycle of births and deaths which terrifies is itself this [his Power of spontaneous self-representation]'. Liberation is just the complete recognition of this fact. Abhinavagupta now concludes the passage in the following way (TA 1.100): bhairava in gurubhir imair anvarthaih samstutah sdstrell 100// Through these [explanations] which are in accordance with what they signify, he is known {samstutah) through [the term] 'Bhairava' by the Gurus throughout sdstra. Here the word anvarthaih seems to mean niruktaih. There is, however, reason to be suspicious of the text here as Jayaratha points out. He comments as follows on the above statement (TAV 1[1] 143,1-11): gurubhih tattacchdstrdvatdrakaih, imaih ebhih samanantaroktaih anvarthaih arthdnugatair vdcakaih, samstutah pancitah sdstre visesdnupdddndt sarvatraiva arthdd uktahl atha ca anvarthaih samyak sakalajagadbharanddisdmarthyapratipddanadvdrena stutah ity arthah, yad uktam 'bhnydt sarvam racayati6* sarvado vydpako 'khile/ iti bhairavasabdasya samtatoccdrandc69 chivahlT

]

^ The editor reads (TAV 1[1] 143,6) ravayati with v.l racayati The form ravayati is problematic as the causative of \rw, but seems to be the accepted reading by Sivopadhyaya (see pp 89 ff. below) and Anandabhatta (VBhK 52,6) They both gloss it by vimrsati 'ideates' which as a technical term m this system of Saivism equally well could gloss racayati, but it is there* ' fore difficult simply to emend Vijnanabhairavatantra 130, tempting though it is, without a proper consultation of the extant manuscripts. There can be no doubt, however, that racayati 'he produces' is Jayaratha's reading and not a corruption, for he quotes the verse also at TA 5.135 (TAV 3[5]488,14—15) and in his commentary there he glosses racayati by antar bahir vd karoti, 'makes internal or external', clearly a reference to Vrac 'produce' 69 So read for satato°\ cf TA 3 136b, VBh 130

84

Indian semantic analysis itiltatha \ . . bharandd bharitasthitih I' nil imaih iti cintyaml gurugaditair iti tu sresthah pdthahll

l

In the verbal codification of the teaching - that is to say, throughout it, because no particular [texts] are specified - he is samstutah, known, spoken of by the Gurus, that is, by those who are responsible for the verbal codification of various doctrines in the tradition, through these expressive terms which have just been mentioned [and which are] anvartha, that is,! which are in accordance with that which they signify. Moreover, [there is another meaning of samstutah:] he is properly (samyak = sam) praised (stutah) [throughout Sastra] by means of these [explanations] which are in accordance with that which they signify since they prove his great Power as the sustainer of the entire universe etc. As stated [in the Vijfidnabhairavatantra]: 'One will become Siva by constantly contemplating in mental utterance the term "Bhairava", understanding that'it means: "he who sustains the universe, he who produces [it], he who bestows all and is pervading the entire universe".'70 Likewise: \ .. his state is all-containing in as much as he sustains the universe.'71 This [reading] 'imaih' is suspect. The best reading is gurugaditaih 'taught/related/ spoken by the Gurus'. As noted by Jayaratha, one carinot be quite happy with the reading imaih here, simply because this is not good Sanskrit. Such an instrumental plural form is found in Agama and in Buddhist Sanskrit,72 but not even there is it common. Now, it is unlikely that Abhinavagupta should slip from the best Sanskrit or make a mistake in grammar. It is therefore more likely that something is wrong with the text at this point. But the reading gurugaditair0 which Jayaratha claims to be the lectio potior is clearly a conjecture since otherwise he would have read it. While this circumstance does vitiate any interpretation of the passage, it seems to me that Abhinavagupta is recapturing his introductory remark anvarthasdstroktaih sabdaih. Moreover, there is fio reason to doubt the reading of the word samstutah. This is taken by Jayaratha to mean 'known' or 'fully/properly praised'. So it is established that Bhairava is known or praised as the referent of the name 'Bhairava' in Sastra, the verbal codification of the teaching in the tradition. Bhairava is thus determined as the referent of his name by fact of his satisfying the property of being known as Bhairava in the tradition, that is, by a certain authorised group of speakers, the
70

72

VBh 130. The reading bhriydt, precative of the active of ^bhr (I bharati), is not adopted by the commentators although it does reflect the notion of bharana or the sustaining of the universe. Sivopadhyaya (eighteenth cent.) reads (VBh 130) bhayd 'by [his] radiance'; cf. p. 89 ff. below. Ananda Bhatta (seventeenth cent.) reads (VBhK on v. 128) bhiyd 'by^fear', moreover, sarvago0 'who moves everywhere' for sarvado0 'who bestows everything'. At TA 5.135 (TAV 71 3(5)488). VBh 24. Sivopadhyaya reads bhantd sthitih. MVUT 2.8; 2.48, 2.58; Edgerton BHSG 21.72.

Praxis: Saiva Kashmir

85

Gurus. And Abhinavagupta is quite explicit on how he understands the term guru (TA 1.235-6): ydvdn asya hi samtdno gurus tdvat sa kfrtitahl samyagjndnamayas ceti svdtmand mucyate tatah II235 // tata eva svasamtdnam jndni tdrayatity adah I yuktydgamdbhydm samsiddham tdvdn eko yato munihll 236// As much as it is his lineage, to that extent it (= the lineage) is recognised as the Guru and as consisting of correct knowledge. Therefore by [the salvation of] that [line] he releases himself. So it is established by reason and scripture that the enlightened 'saves' his spiritual lineage, for the whole is the one master. It is thus evident from Abhinavagupta's own statements that the use of the name 'Bhairava' depends on a certain practice within the Saiva community. Semantics is a very wide social practice founded in the cognitive abilities of individuals which, at least in this case, rests on the authority of the tradition. But, and this is extremely important, one is brought beyond the referential feature to identify the deity as Bhairava through the term itself. An adequate idea of the referent is achieved through analysis of the term 'Bhairava' in which the essence of certain specific ideas is present. The truth-conditions for identifying Bhairava are thus established by means of the nirvacana technique. And it is probably these nirvacanas that are implied by the word imaih 'by those' in the passage.73 According to Jayaratha, the word imaih refers to those just mentioned (samanantarokta), namely those nirvacanas which have just been dealt with. And these nirvacanas are anvartha: they correspond to the meaning and nature of that which the term signifies. It seems worthy of notice that Jayaratha glosses the term anvarthaih by arthdnugatair [vdcakaih '[known] through expressive terms which correspond to [this its] f artha\ It is therefore more than a remote possibility that the word sabdaih \ 'through terms/expressions/linguistic items' of Abhinavagupta's initial stater ment (TA 1.95) includes each full explanation of the name 'Bhairava', the ^explanations through which the deity Bhairava is known. Jayaratha's proposed f variant gurugaditair0 would just mean that the explanations discussed have been sanctioned by the Gurus, the referential performance thus clearly consid> ered a social event in that it relies on a certain practice with regard to its usage • [as well as to its explanation. So the passage makes it clear that the term - 'Bhairava' is used in accordance with the ideas expressed in these explanations which show the equation between the term and that which it signifies, thus taking us from the grammatical to the ontological significance of the name. In nirvacana analysis Abhinavagupta thus finds a powerful means of reinyigorating Bhairava by assimilating Brahmanically acceptable metaphysical conceptions into the idea that the universe is a projection in and of consciousness, the term 'Bhairava' now referring to unconditioned self-representation
73

Cf. the almost parallel statement at TA 1.102; see p. 70 above.

86

Indian semantic analysis

as reality itself. Similarly, nirvacana analysis is for Abhinavagupta an effective means of absorbing the Sakta traditions, the traditions of- Goddessworship, into established forms of Saivism. Abhinavagupta can proceed as he does since in the case of a deity the use of its name rests on some common practice. However, the situation may arise that the reference is - from one point of view or other - regularly "confused, fully or in part. Various features pertaining to various deities or tp different religious practices have become merged so that there exists a practice of using a name but with no single reference. Such a situation may be taken advantage of by a skilled exegete such as Abhinavagupta. By connecting the relevant bits and pieces he is able to establish a coherent practice that suits the religious ideas he propagates. In Abhinavagupta's hands the nirvacana device becomes a powerful means of semantic creation, a device which is ultimately capable of establishing subjective truth as objective truth, that is, as intersubjective truth. This is how a person within a specific culture has the capacity to interpret and determine the contents of thoughts, provided that person has the power to determine the reference of the words of a shared language. And, again, he or she may then have the power to define the meaning of objects and actions, even the power to define others, for this capacity rests not only on a person's ability to be semantically creative, but also on the same person's social position to be so. Ksemaraja's mangalasloka revisited With this background it is possible to return to the interpretation of the mangalasloka from Ksemaraja's commentary on the Vijhdnabhairavatantra. It should be obvious by now that the question of meaning can be properly answered only by answering the more fundamental question of how we come to understand this verse. The evidence is provided by Abhinavagtipta's elaborate analysis of 'Bhairava' and it is only on this basis that the verse by Ksemaraja, Abhinavagupta's pupil, can be appreciated as a powerful statement in religious discourse. A fresh look at the verse reveals the extent to which it is fused with the nirvacanas of the term 'Bhairava' (VBhU 1,2-5): bhirundm abhayaprado bhavabhaydkrandasya hetus tato hrddhdmni prathitas ca bhiravarucdm Tso 'ntakasydntakahl bheram vdyati yah suyoginivahas tasya prabhur bhairavo visvasmin bharanddikrd vijayate vijhdnarupah parah II Bestowing absence of fear for those who are terrified (bhiru) [cf. (2) bhirundm hitakrt], he is the cause of the crying out (dkranda = rava) from fear of transmigratory existence (bhavabhaya) [cf. (4) bhiravasya kdranam], and from that [cry] he becomes manifest in the domain of the heart [cf. (3) bhijanitdd ravdd dhrdi jdtah]; he is the Lord of those who delight in their terrifying roaring [cf. (6) bhiye ravanam ydsdrn td bhiravds tdsdm svdmT], the death of death. Being the Master of

Praxis: Saiva Kashmir

87

that flock of excellent yogins who blow on time (bhera), he is Bhairava [cf. (5) bherah kdlas tarn vayantiti bheravds tesdm svdmi]; agent of sustaining etc. [cf. (1) bharandd ravandc ca; visvam bibharti tena ca bhriyate], he extends his Power throughout the universe, the Supreme whose form is consciousness [i.e., Vijndnabhairava]. All the nirvacanas suggested by Abhinavagupta are contained in this verse. They are, in fact, the very basis of the verse. We see the same formulations in Ksemaraja's commentary on the Svacchandatantra (SvTU 1[1]:3,12-4,7): bhairavo - visvabharanaravanavamanarupah, bhirundm abhayam iti vyutpattyd samsdrindm abhayado 'yam, bhih14 samsdratrdsah tayd janito ravah dkrandah bhiravah tato jdtah taddkrandavatdm sphuritah, asyaiva bhiravasya samsdrabhayavimarsanasydyam saktipdtavasenotthdpakah, bhdni naksatrdni Trayati iti bherah kdlah tarn vayantiti bheravdh - kdlagrdsasamddhirasikdh yoginah tesdm ayam iti dntarah svabhdvah, bhiye pasujanatrdsdya ravah — sabdardsisamutthdkdrddikaldvimarso ydsdm khecarigocaridikcaribhucaricakrarupdndm samviddevindm tdh bhiravah tdsdm ayam svdmi bhairavah, tathd bhairavo bhisanah samsdravighatanaparah, evam dgamesu niruktatvdt, sribrhaspatipddaih sivatandv anvarthavydkhydtasvarupatvdc ca I [And] 'Bhairava' means: 'he whose nature it is (i) to support/nourish the universe // to be supported/nourished by the universe, (ii) to roar [as the transcendental subject that internalises the universe], and (iii) to project75 [the universe so that it appears to be other than him] [cf. (1) bharandd ravandc ca; visvam bibharti tena ca bhriyate]', 'he who bestows absence of fear for those in transmigratory existence', in accordance with the derivation 'security for the fearful' [bhiru- + the taddhita suffix aNby A 4.3.120 tasyedam] [cf. (2) bhirundm hitakrt]; 'he who is born from the fear-cry (bhT-rava)\ the roar, the crying out produced by fear, by terror at [the thought of] transmigration, [that is to say], 'he who becomes manifest to those who cry out at this' [cf. (3) bhijanitdd ravdd dhrdi jdtah]; ['he who produces the fearcry'], he who gives rise to this same cry of fear, [this same] awareness of terror at [the thought of] transmigration, by causing his Power to descend [cf. (4) bhiravasya kdranam]; 'he who belongs to the Bheravas', he who is the inner nature of these Yogins who desiccate bhera, time, the driver of the celestial bodies, [the Yogins] who devote themselves to the trance of the devouring of time [cf. (5) bherah kdlas 5 , tarn vayantiti bheravds tesdm svdmi]; 'he who is the Master of those whose sound terrifies (bhiravdh)\ [i.e., the Master] of the goddesses of consciousness whose nature is the spheres of Khecarl, Gocarl, Dikcarl and Bhucarl, and whose roaring r (-ravah), that is to say, ideation of the elements from A [to KSA] that arise from
74

The text reads samsdrindm abhayadah, bhayam bhih°. Although this is how the text has been taken over by Jayaratha (see p. 76 above), I am suspicious of bhayam before bhih. If it is a gloss on bhih it is both superfluous and in the wrong place. I conjecture that it is a corruption of the * last word of the preceding clause and propose the emendation samsdrindm abhayado 'yam, ^ bhih°. This emendation is supported also by the reading bhirundm abhayaprado0 in the rV mangalasloka quoted above. f5 The element vamana 'emission; projection' does not appear in Abhmavagupta's list, but cf. | the passage by Mahesvarananda discussed p. 88 below. Its inclusion here may betray Ksemaraja's adherence to the Krama.

88

Indian semantic analysis Sabdarasi (the prediscursive matrix of articulate awareness), has the effect of terrifying the uninitiated (pasujanatrdsdya < bhi-) [cf. (6) bhiye ravanam ydsdm td bhiravds tdsdm svdmT]; and also 'Bhairava' means 'the frightening' in as much as his purpose is the destruction of transmigratory existence. [I have explained the words devam and bhairavam at Svacchandatantra 1.1 in this way] because this is how they have been semantically analysed (niruktatvdt) in the scriptures and this is how the nature [of Deva and Bhairava brought out by these terrris] has been explained in accordance with [the nature of] that which they signify (anvartha0) by the revered Brhaspati in his Sivatanu.

This passage too consists entirely of nirvacanas of 'Bhairava', the six known from before, and an additional seventh: tathd bhairavo bhisanah, 'and also "Bhairava" means "the frightening"' which is probably the most obvious lexical meaning of the word bhairava. Interestingly, this last analysis is kept separate from the already established ones by the word tathd and the repetition of the word bhairavah. The passage also makes it evident that there is a tradition in exegesis, a tradition of nirvacana analysis which is founded on Agama and the revered Brhaspati, author of the Sivatanusdstra.76 Indeed, the entire later non-dualistic Saiva tradition relies upon these nirvacanas. The ritual and metaphysical systems developed in Kashmir at the time of Abhinavagupta became powerful in southern India from the eleventh century on, and spread from there all over India. Thus Bhairava appears in reinvigorated form in the same South India where he seemingly enjoyed widespread worship from an early date. Turning to the Mahdrthamafijariparimala, a Pratyabhijiia influenced Krama scripture by Mahesvarananda (fl.c. 1175-1225) of the famous Saiva centre Cidambaram in South India, we meet with the nirvacanas of 'Bhairava' and find that it elaborates on Abhinavagupta's list, adding at the end (MMP 179,25-6): visvam prati bharanaramanavamandndm11 prayojayiteti vd bhairava^ty ucyatel Or, it is he who directs the sustaining (bharana), the resting (ramana) and the emisL sion (vamana) of the universe, hence he is called Bhairava. This nirvacana splits the name into three parts, in accordance with Bhairava's activities of emission (vamana) or projection (srsti), sustaining (bharana) or immersion (sthiti), and the bringing to rest (ramana) or resorption (samhdra) of the universe.78 Mahesvarananda quotes (MMP 180,1-4) the mangalasloka of Ksemaraja, indicating that the bharana-ramana-vamana nirvacana is implied by the expression bharanddikrt 'agent of sustaining etc' in the final line of that verse.
76 77 78

Abhinavagupta too relies on the Sivatanusdstra, see p. 70 above. v 1 vamanam according to the editor This should probably be vamandndm These three phases constitute one of the grounds for the name Tnka (lit. 'triad'), cf. also the tetradic cycle of cognition as found in the Krama, where srsti, sthiti, samhdra and the Nameless Fourth (andkhya) are worshipped in ntual^as the four sequences (krama), the fourth pervading the three See Sanderson 1986 197, cf. TA 1.107.

Praxis: Saiva Kashmir

89

Turning to later Saiva authors, it has already been demonstrated that Jayaratha (fix. 1225-75) expounds the nirvacanas of 'Bhairava' in detail in his commentary on the Tantraloka. Later in the same commentary (TAV 3[5]488) he brings up some individual details pertaining to the nirvacana of 'Bhairava' met with in the Vijhdnabhairavatantra (VBh 130), but these are of no particular interest here.79 Towards the boggling point: analysis through single sounds Having thus shown that the nirvacana device as a means of linking language with metaphysics is an integral and powerful constituent of cultural discourse, let me now turn first to a more extreme form of nirvacana analysis, namely that which is based on the semantics of single sounds, and then to some reservations or objections to the nirvacana device met with in the Kashmirian Saiva tradition itself. A good illustration of this is provided by the lengthy analysis of 'Bhairava' presented by Sivopadhyaya (fix. 1725-75) in his commentary on Vijhdnabhairavatantra 130 (VBhV 113-16). The whole passage deserves close reading, on the one hand simply because it offers an analysis of the sort exhibited above by way of single syllables, and on the other hand because it attacks well-established nirvacanas. For the present purpose I shall however confine myself to certain points of major interest. Sivopadhyaya breaks the term 'Bhairava' down into the four syllables bhd< ai-ra-va. The syllable BHA, which in itself means 'radiance', is identified as samvidrupah prakdsah, 'that effulgence or manifestation which has the form of consciousness'. The syllable AI represents kriydsakti, the Power of action actually its second movement (E-AI-O-AU),S0 when precognitive impulse (icchd, I) and its development (Tsana, I) is englobed within the bliss (dnanda, A) of the ultimate ground (anuttara, A). Sivopadhyaya also quotes the > Ekdksarakosa which identifies AI with Mahesvara, the Supreme Lord. The syllables RA and VA are contained in the verbal form ravayati 'causes to roar'81 which is further explained by vimrsati 'ideates; becomes aware of himself. 'Sivopadhyaya sums this up as follows (VBhV 114,7-9): bhayd samvitprakdsarupayd aikdrah kriydsaktimdn mahesvarah sarvam ravayati vimrsati iti bhairavah in etad eva abhidheyaml Being endowed with the Power of action as the syllable AI by virtue of the radiance (BHA) which is the effulgence of consciousness, the Supreme Lord causes [the X
80 See, however, p 83, note 68 above. See TA 3.94c-104b The causative of ^ru is strictly speaking ravayati, and I strongly suspect that ravayati is a corruption of racayati 'produces' which is the reading of Jayaratha at Vijhdnabhairavatantra • 130; see TAV 1(1)143,6 and 3(5)448,14 The corruption is probably due to the influence of I Abhmavagupta's analyses which involve Vrw, but also the nirvacanas which involve the elements bharana, ravana or ramana, and vamana, cf p. 87 with note 75 and p 88 above, and p. 91 with note 84 below. 79 81

90

Indian semantic analysis universe] to roar (RAVA-yati), that is, ideates (vimrsati) the entire universe: thus [he is called] 'Bhairava'; this alone is what is meant. Sivopadhyaya goes on to give an alternative version of this (ibid.: 114,9-10): yad vd bhayd jndndtmikayd saktyd aikdrena mdhesvaryavyanjikayd kriydsaktyd sarvam vimrsati I Alternatively, he ideates the universe by means of [the syllable] BHAI [his] radiance which is the Power of cognition [and] by the syllable AI which is the Power of action which makes manifest [his] deity.

Next he suggests a nirvacana based on the meaning of Vijndnabhairavatantra 130b: sarvado vydpako 'khile, 'he who bestows all and pervades the entire universe'. This implies the following analysis (ibid.: 114,14): tayd bhayd sarvam subhdsubham rdti daddti sarvam vdti anugacchati vydpakatayd iti bhairavah0 By this [his] radiance [BHA] he bestows [RA: rdti = daddti 'gives'] everything, good and evil, and pervades [VA: vdti = anugacchati vydpakatayd 'he moves along with, m as much as he pervades'] everything, hence [he is called] 'Bhairava'. Here he clearly leans on the authority of the tradition in letting the nirvacana be implied by the two attributes sarvadah and vydpakah. He retains the syllable BHA, but transforms the linguistic items - da 'who gives' and vydpaka 'who pervades' into the phonetically attractive verbal roots Vra and Vva which carry the same meanings. Sivopadhyaya goes on to criticise the reading of Vijndnabhairavatantra 130a which runs bhriydt sarvam rdvayati. Whether this is the reading accepted by Abhinavagupta or not, it is probably known to him as indicated by Abhinavagupta's nirvacana (TA 1.96) based on the roots ^Ibhr and Vrw.82 Sivopadhyaya's view is the following (VBhV 115,10-13): ke cit tu bhriydt sarvam ravayati iti pathanti, tat na sddhu — yadi bhrravah ity ddipadasiddhih sydt tarhi etat samgaccheta I yady api pracurapustakesv api esapdtho } sti, tathdpi dhdranaposandrtho bharatih katamad aksaram apeksya yojanfyah0 Some, however, read [VBh 130a] 'bhriydt sarvam ravayati'. That is not correct. If the original term were established as 'bhrravah', then this would be the suitable construction. Although this reading is found in numerous manuscripts: with regard to which of the [three] syllables is bharatih, which means 'support' and 'nourish', to be joined? The reading bhriydt, precative of the active of ^bhr (bharatih), is thus considered unacceptable, 'although it is found in numerous manuscripts', on the grounds that it would account for the form 'bhrravah* but not ibhairavah\ This warrants the conclusion that Sivopadhyaya considers only such readings accept82

Jayaratha actually quotes VBh 130 with the reading bhnyat See TAV 1(1)143,6, 3(5)448,14, cf pp 83 f with note 70 above Sivopadhyaya prefers the reading bhayd

Praxis: Saiva Kashmir

91

able as would be in accordance with a nirvacana that by some grammatically, or rather phonetically acceptable standard would correspond to the analysed item, 'bhairavah'. The form ravayati, however, does not seem to bother him. Interestingly, he tries out and rejects an explanation which would save the nirvacana that involves the elements bhr and ru. This explanation, he says, relies on the 'crow's eye principle' 83 {bhr ->) Bhaira rava ( — ru) < where the item ra covers both the prior and the posterior. Finally Sivopadhyaya rejects, 'for all its antiquity', the nirvacana which involves the three elements bharana, ravana and vamana, met with in the Svacchandatantroddyota and the MahdrthamanjarTparimalau and discussed above. This nirvacana, he claims, fails to account for the syllable ai. The passage invites a number of observations. According to Sivopadhyaya, a nirvacana has to account for every part of the word that is analysed, otherwise the analysis is not acceptable. Moreover, every part of the word should be infused with meaning. A phonetic link has to hold between the word, or part of the word that is being explained and the explanatory expression. It is also clear that Sivopadhyaya turns to scriptural authority for the ideological foundation of a nirvacana. This ideological foundation pertains to meaning, not to form. For example, the Vijndnabhairavatantra (VBh 130) ascribes to Bhairava the attribute sarvadah 'who bestows all'. This is the verbal codification which provides the semantic basis for claiming that rdti 'gives' is the explanation of the corresponding element ra in bhai-RA-va. This, of course, opens a considerable potential for exegesis, although in the case of Sivopadhyaya it is evident that he sticks to his criteria for determining what is a valid nirvacana and what is not, thus trying to limit the anarchy of infinite possibilities. In this respect it is worthy of notice that he does not fight shy of attacking well-established nirvacanas. One should however note that Sivopadhyaya in no way objects to analysis by means of syllables and single sounds. Such a procedure is in perfect agreement with the statement made by Yaska at least twenty centuries earlier (Nir 2.1): avidyamdne sdmdnye 'py aksaravarnasdmdnydn nirbruydt, 'even when [such] a similarity [with a phonetic change accepted by the grammarians in other cases] is not found, one should analyse on the basis of [a possible] similarity in syllables or in single sounds'.85 Yaska goes on (ibid.): yathdrtham vibhaktih sannamayet, 'the divisions [into syllables and single sounds] should The kdkdksinydya. According to this metaphor, crows are considered to have but a single eyeball which is supposed to roll from the one cavity of the eye to the other as required by the situation 84 Like Ksemaraja, Sivopadhyaya reads (VBhV 115,18) bharanaravanavamandrtho bhairavah, probably intending by ravana the same sense of samhdra 'resorption' Mahesvarananda reads ramana 'the resting' which more obviously gives this sense. 85 Cf pp 37 f. above.
83

92

Indian semantic analysis

be interpreted according to meaning'. This attitude is reflected in the concept of varnasdrupya 'conformity with single sounds' which is frequently resorted to in the exegetical traditions of Saiva Kashmir. Within the domain of mantraanalysis it provides the basic device,86 but it is also employed, as by Sivopadhyaya, in the analysis of ordinary words. As a matter of fact, in his MahdrthamanjarTparimala Mahesvarananda quotes Yaska verbatim (MMP 89,3-4) to defend analysis by means of syllables and single sounds. He also accepts by quotation (MMP 89,3-4) the view proclaimed by Yaska (Nir 2.1): na tv eva na nirbruydt, 'but never indeed should one not analyse'. This, obviously, implies support for analysis by jneans of syllables and single sounds. A clear example of this kind of analysis is found in the explanation of the word ATHA, whose ordinary meaning is approximately 'now; then' (non-temporal). As quoted by Abhinavagupta in his Pardtrimsikdvivarana (PTV 99,9), Somananda (fl.c. 900-50) states: Akdrah siva ity uktas THAkdrah saktir ucyate}1 The linguistic items a and tha are identified as symbolic representations of Siva and Sakti respectively. The context here is the word atha in the fifth sloka of the Pardtrimsikd. Incidentally, Abhinavagupta admits something far stranger here when he suggests that the element tha of atha in the statement athddyds tithayah sarve svardh, '. . . all the 15 vowels beginning with a' may be euphonic (sukhoccdrandrtha) (PTV 101,9). But even a niruktapriyah, lover of the nirvacana device, such as Ksemaraja, feels the need to draw the line somewhere. This is clear from his commentary on Sivasutra 3.29: yo 'vipastho jfidhetus ca. Partly applying Ksemaraja's own interpretation (SSV 115,8 ff.), one may roughly translate: 'He who (yah) resides (°sthah) among the protectors (°pa°) of individuals (avi°, lit. 'animals'), [that is, who controls the Powers thus called,] is indeed (ca) a cause of knowledge (jndhetuh)' But Ksemaraja rejects the following interpretation of the sutra, where the original items are indicated by capital letters: YOgTndro VIjndnaPAdaSTHAh JNAtd HEyas TUcchaH (= visargasaktyd) CA (= implication of kartd [i.e.jndtd kartd ca, 'the agent of knowledge and action']). Since this is rather dense it may be worthwhile to quote the SivasutravimarsinT passage in full (SSV 116,7-118,4): anye tu ' aksarasdrupydt prabruydf iti niruktasthityd 'yo' iti yogindrah, 'vV iti vijhdnam, 'pa* iti padam, lstha* iti padasthah ity asya antyam aksaramI ijna> iti jhdtd, 'he' iti heyah, 'tu' iti tucchaid, visarjamyena visargasaktih, cakarena crnuktasamuccaydrthena karta pardmrsyate ity dsritya, yo yogindro vimarsasaktyd svarupdtmavijhdnapadasthah, sa jhdtd kartd ca avagantavyah; tadd ca asya Cf. Abhinavagupta at TA 5.142. Analysing the pardblja SAUH (cf. Padoux 1963:350 ff.; 1990:416-22) he identifies S with the sensation of the initial cognition of sukha, sitkara, sat, and samya. See also PTV 239,4-7: sakdras tdvat . . . satyasukhasampatsattddindm paramdrthikam vapuh. \ 87 From Somananda's lost vivrti, itself quoting unidentified or lost scripture according to Abhinavagupta who speaks of dgamapradarsanam, 'reference to Agama' (PTV 99,10).
86

Praxis: Saiva Kashmir

93

heyatdm tucchatdm nihsdratdm, na tu upddeyatdm dsddayati iti vydcaksate Ietac ca na nah pratibhdti, paddrthasangater ndticdrutvdt; pratisutram ca Tdrsavydkhydkramasya sahasraso darsayitum saktyatvdtll But others, through the settled rule of the Nirukta that one should analyse [even] on the basis of conformity with [single] syllables, [interpret:] [the element] yo [conforms with] yogindrah 'the great Yogin', vi [with] vijndnam 'knowledge', pa [with] padam 'state'; stha is the last syllable of padasthah 'abides in the state'. [The element] ywa: [conforms with] jfidtd 'the agent of knowledge', he [with] heyah 'to be renounced', tu [with] tucchatd 'vanity', the Power of creative energy (visargasakti) [is implied] by the visarjaniya (h), kartd 'agent' is referred to by the conjunction ca which [here] means that something unstated [but necessary to complete the sense in context] is conjoined. Having resorted to [conformity with syllables in this way], they explain [Sivasutra 3.29 as follows]: 'he (yo°), the great Yogin (yogindrah), who abides in the state of knowledge of the self which is his real nature (svarupdtmavijhdnapadasthah) through the Power of [self-]awareness, should be recognised as [the individual] agent of [both] knowledge and action; and at the same time he realises that this [individual agent of knowledge and action] is to be transcended, is trivial, that is, is without value, that it is not fit to be the object of his aspiration.' But this [interpretation] does not find favour with us since this correlation of terms and meanings is an offence against literary taste. Moreover, since a series of such lines of interpretation could be produced by the thousand for each siltra. Ksemaraja thus clearly objects to such an interpretation, claiming that by this kind of exegesis one could cook up an infinity of interpretations. Yet this reason is second to the specific objection that this particular nirvacana gives an ill-coordinated result as far as meaning is concerned.

The necessity of shared beliefs We have seen how meaning is subject to negotiation through the device of nirvacana analysis. It remains to point out that people were not entirely deceived by it. If meaning is to be socially accessible, and meaning necessarily is, there will have to be a strong element of intersubjectivity involved. Meanings and beliefs are thus two sides of the same coin. It is this intersubjectivity that necessitates the contrast between true and false beliefs. There has to be a certain overlap in beliefs for people to talk about the same things, and we shall now see what befalls the nirvacana device if beliefs are not shared. By way of example I may point to the great eleventh-century Kashmirian Vaisnava satirist Ksemendra who gives some wonderfully derogatory nirvacdnas of the terms diksa and guru, denoting respectively the Tantric ritual of initiation and the teacher who performs it. However, these two elements are the cornerstones of Saivism in practice. For example, Abhinavagupta talks about diksd as follows (TA 1.237-40): tendtra ye codayanti nanu jndndd vimuktatd I diksddikd kriyd ceyam sd katham muktaye bhavetll 237//

94

Indian semantic analysis jhdndtmd seti cej jhdnam yatrastham tarn vimocayetl anyasya mocane vdpi bhavet kim ndsamahjasaml iti te mulatdh ksiptd yat tv atrdnyaih samarthitam I1238 // malo ndma kila dravyam caksuhsthapatalddivat I tadvihantrlkriyd diksd tv anjanddikakarmavat I1239// tat purastdn nisetsydmo yuktydgamavigarhitam I malamdydkarmandm ca darsayisydmahe sthitimll 240// Therefore radically refuted are those who argue that if liberation is caused by knowledge, such actions as initiation {diksd) could not bring release, saying that if the ritual of initiation is knowledge it could still bring about release only of that person in whom it resides (i.e., the Guru) and that there would be^no question of its serving to liberate another. As for the claims of others in this matter that impurity is a substance like a membrane covering the surface of the eye (cataract) and that ritual initiation destroys with an effect analogous to that of antimony (in the case of a cataract), we shall refute them later, being as they are excluded both by reason and revelation; we shall also demonstrate the true nature of mala, mdyd, and karman.

Before turning to the nirvacanas of guru and diksd suggested by Ksemendra, it may be useful to see what prevalent Saiva nirvacanas of the two terms look like. As a traditional nirvacana of the term diksd in non-dualist Saiva sources we may choose the one cited by Ksemaraja in his Svacchandatantroddyota (SvTU3[5]76,17-18): dTyate jhdnasadbhdvah ksiyate pasuvdsand I ddnaksapanasamyuktd diksd teneha kirtitdll Through it knowledge is imparted, and there is destroyed the impression [of contraction] which characterises [the consciousness of] the individual. Therefore, in as much as it is connected with [these activities of] giving and destroying, it is termed 'diksd"' here [in this system]. So here the nirvacana is based on the activities expressed by the roots Add 'give' and Aksi 'destroy', the destruction (ksapana) of the impressions left in the mind being what is intended.88 For a similar nirvacana of the term guru, we may turn to Mahesvarananda's MahdrthamahjarTparimala (MMP 4,10): sa ca guruh grndti prakdsayati visvavyavahdram iti niruktyd sarvanugrdhakah I And he is a guru in as much as he facilitates everything, according to the nirukti: 'he proclaims {grndti), that is, illuminates all practice'. Ksemendra, however, interprets the two terms diksd and guru rather differently in his Desopadesa (Des 8.3):
88

Concerning various nirvacanas of the term diksd common among the Saivas of the Siddhanta, see H. Brunner-Lachaux 1977:3-4. Common to most of them is that they claim diksd to be socalled because it gives the essence of knowledge {^dd 'give') and destroys {^ksi 'destroy') the three impurities etc.

Praxis: Saiva Kashmir gunarahito rutakdri sisyavadhunam sadd gurur gaditah I dindraksayakarandd diksety uktd krtd tenall

95

Being worthless [lit., deprived of qualities, GUna°] and because he makes the wives of his disciples squeal (RUta°), he is called a guru. He performs diksd, so called because it bankrupts you [lit., it causes destruction (KSAya°) of DIndra 'gold (coins)']. Evidently, a strong overlap in beliefs is necessary for the nirvacana device to work properly. A nirvacana is an interplay between properties and names, and thus an attempt to repeat a process of initial baptising through the name itself. Therefore the causal history of a word enters the picture in the sense that a person's concept of a word depends on the kinds of things by which his use of words has been conditioned. This is how the tradition becomes important. The method itself gives legitimacy to a nirvacana analysis, but if a nirvacana can be based on ideas sanctioned by the authorities of the tradition, then its power increases. If a nirvacana has no such basis at all, it serves just to mock; one man's nirvacana may then be another man's con.89 At one extreme, then, it might be held that the range of possible nirvacana analyses is as wide as the range of possible maximal sets of beliefs and that these are as wide as the possible maximal set of sentences consisting of any words beginning with the individual sounds making up the word. Thus^ one reaches a boggling point. But coherence with which beliefs creates a presumption of truth? As noted already, meaning is determined by the community in as much as there is a common language only to the extent that there is a common method of interpretation within a community. Preliminary reflections and conclusions It should by now be clear that a nirvacana analysis is much more than a derivation from a verbal root, and that Yaska's task was different from that of a comparative philologist. Although not every extreme is accepted, it should be evident from the preceding discussion that nirvacana analysis is a powerful device in manipulating the apparently neutral elements of cultural discourse. The obvious question then becomes: how did it come to be so powerful? The above investigation of 'Bhairava' may warrant the following reflections and tentative conclusions. When people share a belief system, then assurance of a change in belief cannot come from outside the system, nor can something inside produce support except as it can be shown to rest on something independently trustworthy. Within a context of shared beliefs, the nirvacana device of interpretation is such a trustworthy entity.
89

It seems natural in this context to refer to the well-known mrvacanas of the names of the four varnas, the classes of the Indian social hierarchy, in the Aggahha-Sutta in the Digha-Nikdya of the Pali canon (sutta 27 in the PTS edition); cf. U. Schneider 1954:578 ff.; A. Mette 1973:33-5; G. v. Simson 1988; R. Gombnch 1992,1992a.

96

Indian semantic analysis

In as much as it is anvartha, the explanatory expression of a nirvacana or nirukti analysis corresponds to the sense and nature of that which is denoted by the term being explained. To determine the reference of a term it is thus sufficient to analyse the term itself. Aiming to persuade the unpersuaded, a nirvacana goes right into the mechanics of a term, pretending to discover meaning in what is already there and seemingly giving the factual affirmation that certain beliefs are true. For example, on the evidence of the nirvacanas of 'Bhairava' presented in the Tantrdloka passage and the other works consulted it is possible to claim that the significant expression of each explanatory unit has the semantic power to identify the nature of a deity such as Bhairava. Whether we can say of any particular that it has necessary or contingent properties depends simply on the way it is described. The important feature of nirvacana analysis is that it is not sufficient to say that a certain property belongs to X, or that a number of properties belong to X, so often or in such ways that people in general believe that X is such that it is recognised by these properties. In nirvacana analysis the attribution of properties happens through the name itself. Now, if this is the case, it is actually possible to argue that the linguistic unit of each explanatory expression is semantically equivalent to the term which is explained, a term such as 'Bhairava' or any singular or general term. The explanatory expressions are thus ideal linguistic forms, arrived at through an analytic procedure accepted by the tradition, and it is really the ideal forms which determine the reference of the term which appears as a linguistic substitute for the ideal one. If, then, these explanations are ideal linguistic forms and if such an ideal form justifies the actual practice of using a term in a specific sense, then the following question becomes crucial: how is it possible that the device of presenting ideal forms of linguistic items - even syllables and single sounds - by means of nirvacana analysis has come to be a hermeneutics *6f the highest esteem? In the universe of discourse only that will count as justification which refers to what is already accepted, and there is no way to get outside the set of shared beliefs and language to find a model for such justification. The answer to this crucial question seems surprisingly simple in that the Indian tradition itself provides us with the model, namely the model of substitution. Here we have, I suggest, a key to a pattern of consistencies within the Indian universe of discourse. Whether substitution may licitly be considered a pattern in Indian thinking in general may be a matter of dispute. It is however settled beyond doubt that it is an established pattern within Brahmanical religious literature ever since the quest for meaning in the fundamental sense started with the Brahmana and Upanisadic literature, later to become technically refined in ritual and grammatical treatises. I put this forward tentatively as a hypothesis to be considered. On this view, a nirvacana works because it is an ideal formulation of the term that is to be explained. The term to be explained is the substitute we meet with in language

Praxis: Saiva Kashmir

97

for a fuller statement with the same semantic content. The nirvacana is a rewording, a rewriting, which is both explanatory and semantically equivalent. Because both expressions are part of the Sanskrit language, in a way nothing new is introduced. 'Bhairava' must mean all these things, because we recognise what the explanations mean and that these meanings do accord with metaphysics. You can give exactly the same truth conditions in a language by making systematic shifts in what the singular terms refer to and what satisfies the predicates. You get a different thought if you get a different sentence to represent it. If you say 'bhairavaK instead of 'yo bibharti\ then it is a necessary truth that the deity referred to is he who sustains the universe. But it is because of the name 'Bhairava' that it is treated as a necessary truth. Now, what challenges does this assumption contain or entail? First of all it becomes crucial to determine the exact relation sustained between a substitute and its corresponding ideal formulation(s), that is to say, the relation between a term and its corresponding explanatory expression. If the above assumptions are valid, this means the Indian conceptions of the relation 'being in the place of. The obvious place to start an investigation of the relationship between term and explanation is Yaska's Nirukta. This is partly because this text laid the foundations of nirvacanasdstra and so established a certain practice, but more importantly because in this text we meet with certain fixed technical methods for expressing a nirvacana analysis. Systematic ways of expression applied over and over again assure us that we are dealing with thought patterns representing implicit models that make up the universe of Yaska, models inherited by the entire subsequent tradition. Secondly, it becomes imperative to investigate the indigenous model of substitution itself, in particular in grammatical literature where it found its most articulate elaboration. Then it remains to see whether the relationship between term and explanation and the interpretation of the explanation as an ideal formulation for which the term met with in the real language is a substitute are compatible with the results of such an investigation.

4 The universe of Yaska

It is evident from the Saiva examples that a nirvacana analysis can be expressed in non-technical language. This is the case also in the Nirukta, for example (Nir 10.22): krstaya iti manusyandmal karmavanto bhavantil vikrstadehd vd9 'krstayah is a name for human beings: they possess action [i.e., they are active] (karmavantah), or: their bodies are stretched out (vi-krstdh)\ Although certain patterns seem to underlie all of Yaska's nirvacanas in the sense that he is establishing a relation between a name and an action or an event, only those explanations which are formulated in a highly technical manner can be expected to give systematic information about the relation sustained between a term and its corresponding explanatory expression. Within the limits of such a bound methodology the Nirukta presents us with five clearly distinguishable ways of analysing words. To facilitate a general presentation, the following constructed examples for the word meghah 'cloud' may be considered: (1) (2) (3) (4) (5) megho mehandt mehatiti meghah megho mehateh megho miheh megho mehatiti satah

Obviously, if we interpret all of these simply to mean 'the word megha is derived from the root miti as has been the habit in modern Indology, the differences between the various ways of expressing an analysis are not accounted for at all. I shall discuss each one of them in turn. The reason for a name: analyses in -andt Yaska explains 109 words in the Nirukta by a construction which involves a verbal noun in -andt. This number is surpassed only by the analyses of the -teh type which is by far the most frequent one. A representative example of the -andt construction is: sindhuh syandandt (Nir 9.26). I suggest the following interpretation: 'sindhu (the river Indus) [is so called] on account of the streaming (syandana)\ A phonetic link is established between the term sindhuh and

The universe of Yaska

99

a verbal element alluded to by the verbal noun syandanam which is to be interpreted as a neuter nomen actionis or bhdvasddhana if we adhere to Indian terminology, that is to say, as 'a streaming'.1 The Sanskrit dictionaries most widely used, record that the verbal nouns in -ana employed in this type of construction are quite frequently attested only in the Nirukta. This warrants the conclusion that these forms are in principle derived by Yaska himself for a technical purpose. The name 'sindhiC is related to the verbal noun by an ablative construction which explains why 'sindhu* came to signify the river of that name, and the construction is thus a natural way of answering the kasmdt 'why?' 2 with which Yaska frequently introduces an analysis. It is also linked to the formula etasmdd eva, 'for the very same reason'.3 A causal interpretation of the ablative is strongly supported by the commentators. For example, Durga remarks on the analysis of vrsabha (Nir 9.22) as follows (D 11:924,3): vrsanat4 retovarsandd dhetoh bhakdram upajanikrtvd vrsabha ity ucyate, 'vrsandt, that is to say, because of the raining of semen, for that reason (hetoh), having made the addition of the syllable bha, he is called vrsabha"? Finally, Durga and Skanda-Mahesvara repeatedly refer to constructions of the -andt type as a hetunirdesa 'explicit statement of the reason'.6 With this background it can be concluded that the -andt construction is a causal explanation revealing the action or event which caused the initial baptising of the river. An analysis of this kind gives no explicit information about how that which is signified is related to the action considered the cause of the name, that is to say, no kdraka analysis is given. In this particular case an interpretation as kartrsddhana is obvious,7 but in most cases we are left to choose among alternative possibilities. This indicates that as a technical analysis it is less precise than several of the other types, a circumstance that is important with regard to the later tradition which finds itself free to choose from the set of possible interpretations that exist within the frames of the model set up by Yaska. As a method of analysis, the -andt construction is older than Yaska. It is met with in the oldest Upanisads, for example in the Brhaddranyaka (5.7.1): viddndd vidyut, 'on account of shattering (viddna) [it is called] vidyut (lightning)'. It is
1

2 3 4 5

6 7

It is difficult to find direct evidence for this m the Nirukta itself since the nouns in question always appear in the ablative case, but the commentators support the above conclusion By way of example, Skanda-Mahesvara consider dhananam to be the nominative of dhanandt (SM 1:5,4), niparanam that of niparandt (SM II 74,9), pananam that of panandt (SM II 95,13), and samnahanam that of samnahandt (SM II 260,14) For this interpretation of kasmdt, see Kahrs 1983 A full discussion of this formula is given pp. 131-52 below. Durga seems to read vrsanat for varsandt as met with in all editions of the Nirukta, only one MS reads varsandt As further examples one may mention Durga's use of tatah and hetoh in his remarks on the analysis of nighantavah (D 1.31,5,16), his statement kim kdranam, 'what is the reason9' before the analysis virodhandt (Nir 6 1), or Skanda-Mahesvara's °hetutvdt (SM II 28,4) See, for example, Durga on visravandt (Nir 5 5) and virohandt (Nir 6 3), cf Kahrs 1980 84-9 Cf. Durga (D II 931,11-13) sindhuh kasmdt I syandandtl sd hy avicchedena visesatah syan data iti, 'why sindhuh^ syandandt for she streams (syandate) in a special way by never ceasing to stream' Note that Durga considers sindhu- to be feminine.

100

Indian semantic analysis

also clear that the -andt type of analysis was established within the nairukta tradition before the time of Yaska. This is evident from the fact that it is the only strictly technical method employed by his predecessor Aupamanyava whenever he is quoted in the Nirukta. An example is (Nir 2.2): dando [. . .] damandd ity aupamanyavah, 'danda (stick; punishment) [. . .] [is so called] "on account of the taming (damanam)", according to Aupamanyava'.8 Evidently a condensed way of expressing typical Brahmana formulations, the -andt construction and the causal way of thinking that underlies it are rewardingly elucidated by a fresh look at the analyses of dpah 'Waters' met with in the Atharvaveda and the Satapathabrdhmana.9 In the Brahmana the analysis appears in the form yad dpnot tasmdd dpah, 'that he got hold of [them], on account of that: dpah\ If in the place of yad dpnot we substitute the verbal noun dpana and put this in the ablative case instead of adding tasmdt, we would arrive at the construction dpandd dpah, 'on account of the gettinghold-of [them, they are called] dpah\ This indicates that in a construction of this kind the verbal noun in -ana is not merely a way of indicating a verbal root. It contains in an ambiguous way the various specific aspects of action which we can attribute to a verb. Not all scholars have interpreted the -andt constructions in the way suggested here. Lakshman Sarup (1921:23) translates the above analysis dando [. ..] damandd ity aupamanyavah (Nir 2.2) as follows: " T h e word (danda) is derived from (the root) dam", says Aupamanyava.' I shall argue why I do not think this is the way to translate a construction of this kind. Yaska's analysis of vrsabha 'bull' at Nirukta 9.22 presents the way of reasoning underlying an -andt analysis quite well: vrsabhah prajdm varsatiti vdl atibrhati reta iti vdl tad varsakarmd varsandd vrsabhah,10 ' vrsabha: [because] he rains offspring or because he emits semen excessively; therefore he is one whose activity (°karmd) is to rain, [and] because of the raining [he is called] l vrsabha'}x The fact that the starting point for the analysis of a word is an action or event related to that which it signifies, makes it possible to suggest alternative analyses for one and the same word according to the action considered to be its foundation. An example is the analysis of devah (Nir 7.15): devo ddndd vdl dipandd vdl dyotandd vd.n But these three alternatives also reveal that an
8

9 11

12

The method is also met with m later texts, for example Visnupurdna 1.5.41: yaksas tu yaksandt, 'as for yaksa: on account of the eating (yaksanam)\ and Abhidharmakosabhasya explaining the word dharma at AK 1.2: nirvacanqm tu svalaksanadhdrandd dharmah, 'as for the nirv cana: on account of10 holding (dhdranam) [its] unique particular, [it is called] a dharma\ [its] See p. 26 above. The context is vrsabhah at Nighantu 5.3. Sarup (1921.146) translates prajdm varsati as 'rains down offspring'. The word prajd can also mean 'semen', but Yaska uses retas for 'semen' in the following sentence. Still, both meanings may be implied. Durga gives support to this (D 11:923,15): vrsabhah kasmdtl sa hi prajam varsati / prajotpattikdranam retah sincati yonau, 'why vrsabhahl Because he rains prajd, th is to say, he emits semen which causes the production of offspring in the female womb.' For further examples, see the analyses of bilvam (Nir 1.14), mustih (Nir 6.1) or pdthah (Nir 6.7).

The universe of Yaska

101

analysis of the -andt type does establish a phonetic link to a verb or a verbal root in as much as the meaning of dipanam and dyotanam is very much the same.13 This raises a question of predominance: is the phonetic link central or is the semantic aspect the primary one? This question finds its answer in the fact that there are instances where there is no trace whatsoever of a phonetic link between the word that is analysed and the verbal noun in -ana. This, I believe, supports the claim that Yaska was primarily concerned with semantics. An example is rasmir yamandt (Nir 2.15).14 Here one would be hard put to trace any phonetic link between the two words whatsoever. Nevertheless, this is one of the examples adduced by S. Varma when he states (1953:29): 'Some etymologies of Yaska shock the philologist on account of their absurdities, as they wildly go against the fundamental laws of phonology.' This, however, is no 'etymology' at all. As noted already, the first three chapters of the Nighantu contain groups of synonyms, and each group is headed by a word which serves to indicate the meaning common to all the words of that particular group. Accordingly, Yaska had no reason to analyse each and every word of the group. He had, on the other hand, every reason to analyse and determine the meaning of the adhikdra-word, and this is indeed his common practice. In the Nighantu (1.5) we find fifteen words grouped under the heading rasmindmdni. The statement rasmir yamandt is made from a purely semantic point of view. Yaska is simply telling us that the word rasmi 'ray; rein' can be used as the adhikdraword of Nighantu 1.5 because of the meaning yamana 'restraining', a meaning common to all the fifteen words of the group. The entire passage of the Nirukta runs (Nir 2.15): rasminamany uttardni pahcadasal rasmir yamandt I tesdm dditah sddhdrandm paficdsvarasmibhih I We are forced to interpret: The following fifteen [words] are names for rasmi 'ray; rein'; rasmi [is the adhikdra-word of the group] on account of [the meaning] yamana 'restraining' [common to all the words of the group]. Among these [fifteen words], the first five are simultaneously denoting horse-reins. Durga has not much to say here, but Skanda-Mahesvara comment as follows (SM 11:90,3-4): rasmir yamanad ity arthakathanam etat, na dhdtupradarsanam,15 'rasmir yamandt: this relates to the meaning and is not a reference to a verbal root'. They go on to give a linguistic analysis (ibid.): raser eva rasmih,
13

Cf. Dhp 1.177 dyuta diptau and 4.42 dipT diptau. Durga too sees no difference m meaning here and remarks (D 1:796,2): dyotanad va dhdtvanyatvam arthaikatvam, '[as for the analysis] dyotanad va there is a different root, [but] the meaning is the same'. For a similar instance, see Durga on the analysis of angam at Nirukta 4.3 (D 1:356,6-7). 14 Other examples are muksijd mocandc cal sayandc cal tatandc ca (Nir 5.19) and vratatir varandc cal sayandc cal tatandc ca (Nir 6.28). The problem has been discussed at greater 15 length by Kahrs 1980:61-76. MSS B and C read dhdtusabdapra0.

102

Indian semantic analysis

'rasmih is indeed of16 rasih (Vra/)'. The Paninlya dhdtupdtha does not list any root ras, but it is met with in Devaraja's Nighantutikd (Mor 1952:52): rasir yamandrtho dhdtuh (sau°), 'rasih is a (sautra) root with the sense of restraining'. Skanda-Mahesvara too operate with a root ras (SM 11:90,4-6):17 rasand rasmir iti katipayaprayogavisaya1* evdyam rasih, rahatyddivat}9 na sarvatra, bandhanapratiter yamanad ity aha - ekatrodakasydparatrdsvdndm I As in the case of [roots] such as rah (rahatih), this [root] ras has only a limited domain of usage [and occurs only in the words] rasand and rasmih, that is to say, not in every instance [of yamana]; but since there is an understanding of [the sense of] binding [for all the words in this section of the Nighantu], he says yamandv, in one context, of water, in another, of horses. What Skanda-Mahesvara indicate here is that Yaska preferred yamandt to rasandt because a root ras would only be acceptable in the cases of rasand and rasmi, not for the other words in this section of the Nighantu. A more straightforward explanation would be that Yaska probably did not know of a root ras. Be this as it may, Skanda-Mahesvara's remark supports the view that rasmi 'ray; rein' can be used as the adhikdra-word of Nighantu 1.5 because of the meaning yamana 'restraining', a meaning common to all the words in this section. This interpretation is supported by an investigation of similar constructions in the context of other adhikdra-woxds in the Nighantu. A good example is the apparently redundant analysis stotd stavandt (Nir 3.19). The grammatical
16 17

This is a preliminary and neutral way of interpreting genitives of the -eh and -teh kinds; such analyses will be discussed in detail further on. The unddi analysis is by adding the suffix mi to as which is replaced by ras; so, for example the DasapadT (DPU 1.15): ase asa ca. The Vrtti comments: aser dhdtor mipratyayo bhava rasa ity ayam ddesahl asu vydptau sau°, asa bhojane krai° I asnute asnatiti vd rasmi mayukahl karttd, 'the suffix mi is added after the root as, [and] this [element] ras is [its] substitute; the fifth class root as when the sense of attaining [is to be denoted] or the ninth class root as when the sense of consuming [is to be denoted]: it attains or it consumes, hence rasmi, i.e., a ray of light; [it denotes] the agent [of the actions denoted by the two roots as]\ Ujjvaladatta comments on PPU 4.46 as follows: asu vydptau// ato mihl dhdto rasddesahl rasmih kirano raj jus ca, '[the root] as [occurs] when [the meaning] vydpti [is, to be denoted]; [the suffix] -mi [is added] after it; the substitute ras [occurs] in the place of the root, [which gives you] rasmi [in the senses of] ray of light and rope/rein'. A similar interpretation is met with in Hemacandra's Unddisutra 688 (Kirste 1895:116): aso ras cddih, '[the suffix -mi (687) is added] after [the root] as and r [occurs] as the beginning'. Hemacandra himself comments (ibid.): asauti vydptau I ity asmdn mih pratyayo [bhavati] rephas ca dhdtor ddir bha rasmih pragraho mayukhas ca, '[the root] as [occurs] when [the sense of] vydpti [is to be denoted], [and] after it the suffix -mi [is added] and r appears as the beginning of the root, [which gives you] rasmi [in the senses of] rein and ray of light'. Further information is yielded by Hemacandra's analysis of the word rasand (Kirste 1895:47): aso ras cddau, '[the suffix -ana is added] after [the root] as and r [occurs] at the beginning'. In his commentary Hemacandra remarks (ibid.): rasim eke prakrtim upadisanti sd ca rdsi rasand rasmi ity a prayujyata ity dhuh, 'some teach rasih as the stem, and say that it is used here: rasih, rasand, [and] rasmih\ 18 MSS B and C read katiprayaprayoga0. The passage occurs in Devarajayajvan's commentary on the Nighantu, and the edition of Mor (1952:52) reads katipataprayoga0. 19 MS D reads haratyddi0 and Devaraja (ibid.) reads bharatyddi0.

The universe of Yaska

103

formation of stotr is so completely regular that no further analysis seems to be called for. In Paninian terms, the krt (A 3.1.93 krd atin) suffixes (A 3.1.1 pratyayah) NvuL (with -aka replacing NvuL by A 7.1.1 yovor andkau) and trC (-tr) (A 3.1.133 nvultrcau) are added after (A 3.1.2 paras ca) a verbal root (A 3.1.91 dhdtoh) to denote the agent (A 3.4.67 kartari krt). The wording stotrndmdny uttardni trayodasal stotd stavandt (Nir 3.19) shows to better advantage if we interpret: 'the following thirteen [words] are names for stotr "one who praises"; stotd [is the adhikdra-word of Nighantu 3.16] on account of [the meaning] stavana "praising" [common to all the words of the group]'. Just like rasmi in Nighantu 1.5 the word stotr 'praiser' is the suitable adhikdraword of Nighantu 3.19 because of the meaning stavana 'praising' common to all the words in this group. To conclude, it is evident from this that constructions of the -andt type can refer exclusively to the semantic content of a word. This important fact also explains why such constructions occur as additional semantic elucidations to analyses of the -teh type.20 There is accordingly every reason to be cautious even when a phonetic link seems to be established between a term and the verbal noun in -ana. To regard these constructions in the Nirukta as first and foremost grammatical analyses is apt to lead one astray. They are causal explanations revealing the action or event which caused the initial baptising of that which is signified by a term. Another way of giving a reason: mehatiti Yaska has one more way of giving a causal explanation for why something is called what it is called, exemplified above as mehatiti meghah, '[being characterised by the description] "it rains (mehati)" [he is called] megha\ The thought pattern underlying it is also to a certain extent identical with the one underlying the analysis of dpah 'Waters' met with in the Atharvaveda 3.13.2c-d: tad dpnod indro vo yatis tdsmdd apo dnu sthana, 'then Indra got hold of you as you flowed, hence you are [called] Waters thereafter'.21 This is a more complex causal explanation in as much as the author has made it clear just how the Waters participated in the action considered the cause of their name. And, noteworthy, the narrative is in the past tense, thus referring to one specific event. The Nairuktas in general dissolve historicity, and use the present tense instead. In other words, an analysis of the type mehatiti meghah is an instance of expressing a familiar karate-analysis. Let me briefly return to the examples of kdraka-an&lysis in the Nirukta adduced previously22 and reconsider them as examples of a particular type of nirvacana analysis. Yaska analyses dhanvan (dhdnva, Nigh 4.2) as follows (Nir 5.5): dhanvdntariksaml dhanvanty asmdddpah, 'dhanvan, i.e. antariksam (the
20 21

See Kahrs 1980:110-24; by way of example: disah kasmdtl disatehl dsadandtl api vdbhydsandt (Nir 2.15); pdsyd pdsasamuhah I pdsah pdsayatehl vipdsandt (Nir 4.2). 22 See p. 26 above. See p. 54 above.'

104

Indian semantic analysis

?

intermediate space between heaven and earth; atmosphere), [is so called because something] flows (dhanvanti) from it, [namely] water'. Yaska thus suggests an interpretation as apdddnasddhana: the atmosphere is that from which water flows, and because of that flowing it is called dhanva. Another example is the analysis of ahar 'day' (Nir 2.20): ahah kasmdtl updharanty asmin karmdni, 'why aharl [because] one carries out (updharanti: upa-d-Ahf) [something] in the course of it, [namely] actions'. Yaska gives an analysis as adhikaranasddhana: & day is that in which actions are carried out, and because of that carrying out, it is called ahar. Likewise the word grisma 'summer' (Nir 4.27): grismo grasyante 'smin rasdh, 'gnsmah [is so called because something] is devoured {grasyante) during it, [namely] juices'. Yaska gives an interpretation as adhikaranasddhana: the summer is the season during which all juices are devoured, and because of that devouring it is called grisma. It is clear that an analysis of the type mehatiti meghah gives a causal explanation and also an interpretation with regard to kdraka. It is in the latter respect that it differs from an analysis of the -andt type which does not specify how that which is signified participates in the action considered the cause of the name. As for the relationship between a term and its corresponding explanatory expression, it is also in this case a causal one. Analyses in -teh and -eh It is now time to turn to the method exemplified as megho mehateh. The first thing one has to face is that mehateh terminates in a case form which is not immediately clear. The method presents the 3rd sg. present active form of the verbal root Amih treated as a nominal /-stem to make it declinable. So far in modern Indology the ending -eh has been interpreted as an ablative ending, the interpretation 'derived from' being taken for granted. Without excluding the ablative possibility - although the ablative would definitely have to be interpreted differently from what has been common so far - 1 shall in what follows present evidence which suggests that it is a genitive ending. This would open up for far more interesting interpretations, in particular with regard to the relation between the word analysed and the expression used to analyse it. The mehateh type is by far the most common technical method of analysis met with in the Nirukta. In a very few instances, however, the analysis is given in the form megho miheh. In this case the verbal root is turned directly into a noun by the addition of the nominal suffix -/. Both of these constructions, mehatih and mihih, are sanctioned by Katyayana as valid ways of referring to a verbal root.23 With this background, let me now turn to the type of nirvacana expressed by means of a formal analysis in -teh and what it entails in more detail. Because most of the -teh material from the Nirukta adds little to our know23

A 3 3 108, vt 2 ikstipau dhdtunirdese

The universe of Yaska

105

ledge about this type of analysis as such, I see no point in discussing this material in extenso. Instead, as a sample of what we are dealing with, I shall begin by presenting the first fourteen analyses occurring in the text, together with the relevant remarks by the commentators. I shall then proceed to discuss in detail the possibly interesting cases. In what follows the ending -teh has been translated very neutrally as 'of, denoting the relation of 'belonging to', under its aspect as a possible genitive ending, and as 'from', denoting th$ relation 'being from', under its aspect as a possible ablative ending. The analyses are presented in the same order as they occur in the Nirukta, and numbered by reference to the chapter and passage in which they occur. If further distinction is required, a letter is added after the paragraph number. (1.4a) vaydh sdkhd veteh I vdtdyand bhavantil vaydh, [meaning] branches, is of/from vetih (A/VF 'move'). They are such as move m the wind. This refers to RV 6.24.3 which exhibits the form vaydh (pi.), thus incidentally analysed in this context. Skanda-Mahesvara annex the attribute gatyarthasya to veteh, hence considering it a genitive form (SM 1:51,9): vayah sakha veteh gatyarthasya. Similarly, Nllakantha (NSV 1.2.55): gatyarthasya vayah sakha veter dhdtor uddhrtdh.24 Nllakantha here adds the word dhdtoh, genitive singular of dhdtu- 'verbal root'. To determine the exact sense of the word uddhrtdh could be of some importance for understanding Nllakantha's way of interpreting an analysis of this kind. It seems to me most likely that the word carries the general sense of 'related; declared; said to be'. The phrase vdtdyand bhavanti I would consider an additional explanation which states the reason for choosing the activity expressed by the root Vvf as the basis for the analysis of vaydh which reasonably lends itself to an interpretation as kartrsddhana, that is, as expressing the notion of agency with regard to the activity denoted by the verb vetih: Vvf 'move'. Skanda-Mahesvara, on the other hand, take vdtdyand bhavanti to be a separate analysis of vaydh25 (SM 1:51,11): vatayana bhavantln dvitiyam nirvacanam, vdto 'yano ydsdm vdtena yds cdlyanta ity arthahl 'vdtdyand bhavantV is a second analysis: those for which the wind is a mover, that is to say, those which are caused to move by the wind. This indicates a karmasddhana interpretation of vaydh, more in accordance with the meaning of the noun vdtdyanam (neuter gender) met with in classical According to the editor, the MS reads veteddhdtoruddhrtam. So also NSV 12.56, however the lacuna in the text is filled up with the relevant words from the commentary of Skanda-Mahesvara The editor may be justified in doing so since v Nllakantha obviously relied on this commentary, cf in this respect Vijayapalah 1982 46 ff (upodghdtah).
24 25

106

Indian semantic analysis

Sanskrit, namely, 'an opening in the wall through which the wind can pass; airhole; window', though this strictly speaking would require adhikaranasddhana. The word vaydh could thus be understood to denote branches in the sense of 'such through which the wind moves/such as are moved by the wind', maybe involving the element va- from vdta 'wind', that is to say; from Vva26 'blow', and -ya from Vi 'move'. Durga does not cite or comment upon the words vdtdyand bhavanti which may therefore be a later addition. Sayana quotes the entire passage at RV 1.59.1. (1.4b) sdkhdh khasaydhl saknoter va/ sdkhdh, that is, khasaydh 'such as abide in the air'; or it is of/from saknotih i^isak 4 be able'). The word sdkhdh is analysed here since it is the synonym given for vaydh (cf. 1.4a). The analysis saknoteh is cited by Skanda-Mahesvara and Nllakantha (NSV 1.2.58). It is repeated and once more recognised by Skanda-Mahesvara at Nir 6.32.27 Durga, on the other hand, ignores the analysis altogether. (1.6) cittam cetatehl cittam 'mind; thinking faculty' is of/from cetatih (^cit 'perceive; reflect'). The context is RV 1.170.1. The analysis is not quoted in SkandaMahesvara's commentary where, iti fact, the latter part of Nir 1.6 is omitted or missing. It may be noted, however, that the analysis is included by Nllakantha28 who generally follows Skanda-Mahesvara.29 Durga suggests karanasddhana (D 1:77,16): cetaty anendrthdn iti cittam, 'one perceives [something] by means of it, [namely,] things/meanings, hence cittam - means of perceiving'. (1.7a) magham iti dhanandmadheyam I mamhater ddnakarmanahl

mag ham, being a name for dhanam 'wealth', is of/from mamhatih (^mamh) which denotes (lit.: has) the activity of giving. The context is RV 2.11.21. The word magham is listed among dhanandmdni Nigh 2.10, and the finite verb mamhate among ddnakarmdnah Nigh 3.20. The commentators suggest karmasddhana, for example, Durga (D 1:79,13): diyate hi tat, 'for it is given'.30
26

Shortened to va by A 3.2.3 dto 'nupasarge kah which teaches that the krt suffix Ka (-a) is added after a verbal root which ends in a when it is not preceded by a preverb but co-occurs with a noun 27 denotes its direct object. This accounts for forms such as goda- 'giver of that cow(s)'. The NSV breaks off at Nirukta 6.17. 28 NSV, unnumbered sloka concluding the second pdda of the first adhydya: cittam vijndnam evdha cetater jhdnakarmanah. The gloss vijndnam may indicate bhdvasddhana although th word may lend itself29 almost any kdraka analysis, an interpretation as karanasddhana being to quite common. Vijayapalah 1982:46 ff. (upodghdtah). 30 Cf. Skanda-Mahesvara (SM 1:68*14): tad dhi diyate; Nllakantha (NSV 1.3tl0). mamhyate diyate tad dhi tena tan magham ucyate, 'for it is bestowed (mamhyate), that is to say, given; hence it is called magham'.

The universe ofYaska (1.7b) daksind daksateh samardhayatikarmanah I vyrddham samardhayatiti I

107

daksind '[rite of] the ceremonial reward(s)'31 is of/from daksatih (^daks) which denotes the activity of fulfilling: it fulfils that which is unfulfilled.32 The context is RV 2.11.21. Durga (D 1:79,14) substitutes samardhayatyarthasya for samardhayatikarmanah which he thus considers a genitive form. Whether or not Yaska suggests an analysis as hetu, causal agent, in choosing the causative samardhayati is hard to tell. If not, an interpretation as kartrsddhana seems reasonable. (1.7c) daksino hastahl daksater utsdhakarmanah Iddsater vd sydd ddnakarmanah I daksinah 'the right' - that is, the hand - is of/from daksatih (^daks) which denotes the activity of being strong; or it may be of/from dasatih (^Idds) which denotes the activity of giving. The analysis is given incidentally in the context of daksind (1.7b). In the commentary of Durga utsdhakarmanah is replaced by utsdhdrthasya (D 1:79,19) and ddnakarmanah by ddndrthasya (D 1:80,1), both forms thus considered genitives. The first analysis suggests an interpretation as kartrsddhana (utsahate 'acts with strength'), while the second implies karanasddhana. So Durga (D 1:80,1): tenaiva hiprdyena diyate, 'for with this alone is [something] usually given'; Skanda-Mahesvara (SM 1:69,6): tena hi diyate yad deyam, 'for with this is given what is to be given'.33 (1.7d) hasto hanteh/ hastah 'hand' is of/from hantih (^han 'hit; slay'). The analysis comes up through the preceding analysis of daksinah 'right hand (hastahy (1.7c). Durga suggests an interpretation as karanasddhana (D 1:80,2): tena hi hanyate, 'for with it [something] is struck'. (1.7e) bhago bhajatehl bhagah 'share; happiness' is of/from bhajatih (^bhaj 'divide; share; enjoy'). The context is RV 2.11.21. The analysis is not cited by Durga or SkandaMahesvara. Nilakantha, on the other hand, quotes it and interprets it as
31

On the complex notion of the daksind, see Malamoud 1976; cf. review by Heesterman (1979); further Gonda 1955:75-6. Malamoud (1976.173, note 3) also deals with the present Nirukta passage and its continuation. See in this respect also Kahrs 1980:110-13. 32 To fit the context of RV 2.11.21; otherwise perhaps: 'it makes successful that which is unsuccessful'; cf. Durga (D 1:79,14): yajhe hi yat kirn cid vigatardhikam bhavati tad lyam samar ; dhayati visistam yajnasya sddhanam etad iti, 'for whatever is deprived of (vigata = vz) succes at the sacrifice, that it makes successful, it is a particular implement of the sacrifice'. 33 NUakantha follows Skanda-Mahesvara (NSV 1.3.19): diyate tena yad deyam na savyena nisedhatah

108

Indian semantic analysis

karmasddhana (NSV 1.3.21): sevyate34 sa sukhdrthibhih, 'it is worshipped/ enjoyed by those who strive for pleasure'. It could even be the deity Bhaga that Nllakantha has in mind.The analysis bhajateh appears Nir 3.16 (cf. 3.16a). (1.7f) viro virayaty amitrdnl veter vd sydd gatikarmanah vlrayeter vdl virah 'hero': he agitates/scatters {vlrayatih wvilr)35 enemies; or it is of/from vetih ( vvf) which denotes the activity of moving; or else it is of/from vlrayatih (yvlr\ viray 'be heroic').36 The context is RV 2.11.21, suvirdh. Durga (D 1:80,5) explains gatikarmanah by gatyarthe vartamdnasya, '[of vetih] which occurs in the sense of going', clearly a genitive interpretation. Moreover, he adds the genitive attribute vikramdrthasya to viray ateh (D 1:80,6); cf. Dhp 10.355 vira vikrdntau, '[the root] vir [occurs] when courage [is to be denoted]'. Nllakantha gives the genitive gatyarthasya for gatikarmanah (NSV 1.3.26): gatyarthasydtha vd veter amitrdn prati gacchati31 'or else it is of vetih which has the sense of moving: he moves against the foes'. This interpretation as kartrsddhana seems reasonable also with regard to the other explanations.38 (1.8a) gdyatram gdyateh stutikarmanah I gdyatram (ritual chanting in gdyatrl metre) is of/from gdyatih (^Igai) which denotes the activity of praising.39 The context is RV 10.71.11. Durga (D 1:84,10) replaces stutikarmanah by the unambiguous genitive stutyarthasya and goes on to suggest an interpretation of gdyatram as karanasddhana:40 stuyate hy anena, 'for something is praised with it'.
34 35

Cf. Dhp 1.1047: bhaja sevdydm. j Sarup (1921:10, note 7) makes the following unjustified remark: 'Durga paraphrases Vlrayati by ndndprakdram mdmyati, i.e. "he kills in various ways". He seems to take vir as a noncompound root, and is supported in this interpretation by Dhatupdtha, xxxv,49, where vir is enumerated as a verb of the tenth class. But Yaska appears to take it as a compound of vi + Ir (to disperse), for he distinguishes it from the denominative verb, cf. his third derivation.' Sarup is probably misled by the reading found in the edition by Satyavrat Samasraml (1882-91). The reading mdmyati is only met with there, while all other editions or MSS of Durga's commentary read ndndprakdram Trayatity arthah, 'that is to say, in various ways (^vi) he agitates'. Sarup is of course right when noting that Yaska's first analysis vlrayati is of vi + ^lin 36 pw gives: 'viray, vlrayate sich mdnnlich -, sich tapfer beweisen. Act. einer Etymologie bewdltigen.' The assumed parasmaipada is probably based on this very analysis; cf. MW '2. vir, cl.lO.A. (Dhatup. xxxv,49; rather Nom. fr. vird below) virdyate, to be powerful or valiant, display heroism, RV.; VS.; TBr.; (P. vlrayati) to overpower, subdue, Nir. i.7.\ But there is no reason to establish lexically a parasmaipada form vlrayati on the basis of this passage since Yaska pays no heed to the circumstance that a verb may occur only in the dtmanepada when he refers to it as an /-stem in the masculine as, for example, vlrayatih. 37 Cf. Skanda-Mahesvara (SM 1:69,10): gacchaty amitrdn prati. 38 Cf. Durga, note 35 above, and Skanda-Mahesvara (SM 1:69,9): vividham Irayati prerayati, 'variously (vividham = vi) he agitates, that is, sets in motion'; cf. Dhp 2.8 Ira gatau kampane ca. 39 Yaska thus gives gdyatih stutikarmd. The Nighantu lists gdyati and stduti under arcatik rmdnah (Nigh 3.14). The Paninlya dhatupdtha gives gai sabde in the bhvddigana (Dhp 1.965), while the meaning entry40 stutau is met with in the juhotyddigana (Dhp 3.25), yielding the gd present form jigati. Similarly Skanda-Mahesvara (SM 1:73,1).

The universe ofYaska (1.8b) sakvarya rcahl saknotehl sakvaryah, [meaning] Stanzas, is of/from saknotih {^sak 'be able').

109

The context is RV 10.71.11 which exhibits the form sakvansu, locative plural of sakvarT-. The Nighantu, however, includes sakvarT among bdhundmdni (Nigh 2.4), evidently as the nominative dual of a stem sakvari-41 and also among gondmdni (Nigh 2.11) where it appears to be the nominative singular of a stem sakvarT-.42 Yaska himself suggests karanasddhana: tad yad dbhir vrtram asakad dhantum tac chakvarmdm sakvantvam, 'that he [= Indra] was able to slay Vrtra by means of these, that is the sakvarT-ness of the sakvaryah'.43 The analysis is not met with in the commentary of Skanda-Mahesvara. (1.9a) aksi casteh I anakter ity dgrdyanahl aksi 'eye' is of/from castih (^caks 'see'); it is of/from anaktih {^ahj 'reveal; make appear', according to Agrayana. The context is RV 10.71.7, aksanvantah. Durga and Skanda-Mahesvara interpret casteh and anakteh as genitive forms: Durga adds the attribute pasyatyarthasya to casteh (D 1:87,18), while darsandrthasya is preferred by Skanda-Mahesvara (SM I:75,7);44 to anakteh is annexed the attribute vyaktyarthasya in both commentaries (D 1:87,19; SM 1:75,7), Skanda-Mahesvara employing a more complex way of expressing the information considered necessary: anakteh - 'ahju vyaktimraksanagatisu' ity asya vyaktyarthasydksity dgrdyanah, 'according to Agrayana aksi belongs to anaktih, that is, to that concerning which it has been stated [Dhp 7.21]: "[the root] anj [occurs] when appearing, anointing or moving are to be denoted", that is, to that which has the sense of appearance'. Under the first alternative, an interpretation as karanasddhana is suggested ?by the commentators, for example by Durga (D 1:87,18): drsyate hy anena, 'for .[something] is seen by means of it'.45 Under the second alternative, however, Jit seems reasonable to interpret it as kartrsddhana (NSV 1.3.113: anakti . . . tyyaktikaroti 'it makes manifest') or as karmasddhana (NSV 1.1.114: svayarn vd vyaktam anyebhyo bhrsam tejasvi projjvalam, 'or it is itself made manifest,

)

f Cf. Devaraja Yajvan (ed. Mor 1952:1:207): saknutah karmdni karttum, 'those two are able to perform activities'. f Cf. Devaraja Yajvan (ed. Mor 1952:1:246): saknoti0, 'it is able ...'. f It seems reasonable to assume that this is a quotation from some Brahmana text, although I : , have not been able to trace the source. Sarup (1927) refers to Kausitakibrahmana 23.2 and >" Aitareyabrdhmana 5.7.3 where similar passages are met with. 44 NSV 1.3.112: caster darsanakarmanah, probably for metrical reasons since Nflakantha in ^; general follows the commentary of Skanda-Mahesvara. 5 f Cf. Skanda-Mahesvara (SM 1:75,7): drsyate hi tena\ Sayana at RV 10.71.7: anena drsyate sarvam, 'by means of it everything is seen'; Nilakantha (NSV 1.3.113): caste 'nena, '[some/ thing] becomes visible / one looks at [something] by means of it'. Patanjali (Mbh II: 119,19-20) ^ gives an analysis as karanasddhana by asnotih (^las 'reach; obtain') with the unddi-suffix si asnoter ayam aunddikah karanasddhanah sipratyayahl asnute 'nenety aksi II cf. howev Bhdsya at rule 8.2.48, vt. 3 (Mbh 111:408,21 ff.).

110

Indian semantic analysis

being vehemently [more] brilliant, that is, luminous, than any other [parts of the body]'; cf. below). Sarup (1921:12) takes ^anj in the sense of 'be beautiful'. This is possible,46 but I prefer the sense of 'cause to appear; reveal'. Durga comments in the following way (D 1:88,4-6): ete aksini itarebhyo 'ngebhyo vyaktatare spastatare prakdsatare bhavatah andhakare 'pi hy ete prakdsete naktamcarddindm na tathetardny angdm° The two eyes are more revealed, that is, more clear, more visible, than any other parts of the body, for even in the dark these two become visible, that is, of nightcreatures etc.; not so the other parts of the body. Similarly, Skanda-Mahesvara (SM 1:75,8): taddhy ujjvalatvddvyaktataram anyebhyo 'ngebhyah, 'for it [=the eye], because it is luminous, is more visible than any other parts of the body'. (1.9b) karnah krntateh/ nikrttadvdro bhavati I rcchater ity dgrdyanahl karnah 'ear' is of/from krntatih (^Ikrt 'cut'): its aperture has been cut into it; it is of/from rcchatih (Vrcc/i/r 'move'), according to Agrayana. The context is RV 10.71.7, karnavantah. That the form krntateh is considered a genitive by Skanda-Mahesvara is evident from the statement (SM 1:75,9): karnah krntateh 'krti chedane* ity asya, '"karnah is of krntatih", that is, of that concerning which it has been stated [Dhp 6.141]: "[the root] krt [occurs] when cutting [is to be denoted]"'. The remarks by Skanda-Mahesvara on the analysis rcchateh are worthy of notice (SM 1:75,11): artter gatyarthasya 'pdghrddhmdstha* ity rcchddesavidhdndd rcchatir gatyarthah I tasya karna ity dgrdyanah I Since, according to [A 7.3.78] pdghrddhmdsthd etc., the substitute rcchA is taught for arttihlr which has the sense of movement, rcchatih has the meaning movement. [The word] karnah is of this [tasya = rcchater gatyarthasya], according to Agrayana. Here rcchateh is clearly considered a genitive form. Moreover, the passage demonstrates how meaning is involved when one linguistic item is replaced by another, since here it is argued that rcchatih is gatyarthah by force of its replacing arttih which is listed as gatyarthah, 'having the sense of movement', in the Dhdtupdtha: rccha gatTndriyapralayamurtibhdvesu (Dhp 6.15), further r gatiprdpanayoh (Dhp 1.983) and r gatau (Dhp 3.16).47 The meaning gati is common to all the meaning-entries.
46

Several dhdtupdthas include kdnti 'beauty' in their meaning-entries for Va/z/, for example the MddhavTya Dhdtuvrtti (ed Dwarikadas Shastri 1964 504) anju vyaktimraksaiiakdntigati Likewise the Kdsakrtsna-dhdtuvydkhydnam (ed Yudhisthira Mlmamsaka 1965/6a:174). See 47 also Bhaglratha Prasada Tnpathl 1965.7. Cf. Nir 2.25. artter gatikarmanah.

The universe of Yaska

111

With regard to kdraka, it seems reasonable to interpret it as adhikaranasddhana under the first alternative, since Yaska states that an aperture has been cut into it, nikrttadvdro bhavati^ Under the second alternative, karnah lends itself to an interpretation as kartrsddhana on the basis of the untraced quotation offered by Yaska here: rcchantiva khe udagantdm, 'moving, as it were, the two ears went up'. 49 (1.9c) dsyam asyateh/ dsyandata enad annam iti vdl dsyam 'mouth' is of/from asyatih (^as 'throw'); or: food is streaming (d^syand) towards it, thus [dsyam]. The context is RV 10.71.7, ddaghndsah 'which reach to the mouth'. Yaska takes a in the sense of dsyam.50 The genitive attributes ksepdrthasya and ksepandrthasya are attached to asyateh by Durga (D 1:89,4) and SkandaMahesvara (SM 1:76,9) respectively. The analysis asyateh opens up for a number of different kdrakainterpretations. Durga suggests adhikaranasddhana (ibid.): ksipyate hy etaddbhimukhyendnnam, 'for [something] is thrown towards it, [namely] food'. Nllakantha, following Skanda-Mahesvara,51 proposes adhikaranasddhana (NSV 1.3.121): ksipyate hy annam asminn ity dsyam asyateh, 'food is indeed thrown into it, hence [the analysis] dsyam asyateh'. The wording here is worthy of notice since it illustrates beyond doubt that it is the whole analysis dsyam asyateh that is justified by the faara&a-interpretation, not only the word dsyam. Finally, Skanda-Mahesvara offer an alternative analysis as kartrsddhana (SM 1:76,9): ksipati vd varndn, 'or, it throws [out] speech-sounds'. The material presented so far represents a fair sample of analyses in -teh, all together the first eighteen occurrences in the Nirukta, including alternative analyses. From the material of the Nirukta presented so far, it is difficult to determine whether the forms in -teh are genitives or ablatives. With a few exceptions which will be dealt with in detail, this is the situation throughout the Nirukta. The only attribute to forms in -teh which occurs regularly is °karmanah, & bahuvrThi compound in the masculine used to indicate the meaning of a verb, or, more precisely, the action referred to by a verb. This
48

Cf Skanda-Mahesvara (SM 1.75,10) sa hi . chinnadvdro garbhdvasthdydm kendpi krtabilah, 'for it is something which has its aperture cut, that is, at the embryonic stage an " opening has been made by some [power of fate]' Note that kha (n.) which carries the sense of * ,49 'ear', basically means 'hollow; apeiture' Sarup (1921 12) translates* 'Going upwards, as it were, they have protruded in space' He thus * considers khe to be the locative singular of kha (n) 'space', although the lack of proper sandhi ^ suggests that khe is nominative dual of kha (n ) m the sense of 'ear', the dual ending -e being xs pragrhya, uncombmable The point is discussed at length by Skanda-Mahesvara (SM 1:75,13-76,8) who conclude as I have suggested m my translation See also NSV 1 3 117-20 > The close association of sound with space is a common feature m Indian thought Samkhya, for example, considers dkdsa 'space' to be the substratum of sound |° Cf. Sayana at RV 10.71 7 dsyasabdasya prsodardditvdd dkarddesah, 'since the word dsya i 4- of the prsodarddi-type, the syllable a occurs as its replacement'. 51 SM 176,9: ksipyate hi tatrdnnam

112

Indian semantic analysis

leaves us with the same problem as before: has °karmanah a genitive or an ablative case ending? The situation changes drastically when we turn to the commentaries. We have no guarantee, of course, that the interpretations offered by the commentators represent the view of Yaska. If they are all in agreement, however, it would not be unreasonable to assume that they have kept a tradition. It is therefore noteworthy that when they make their position known, they invariably treat the forms in -teh as genitives. Durga quotes or comments upon sixteen of the eighteen forms in -teh; nine of these are clearly taken to be genitives, none as ablatives. SkandaMahesvara quote or comment upon fifteen; six of these are clearly interpreted as genitives, none as ablatives. Nllakantha quotes all eighteen, clearly taking two of them as genitive forms, none as ablative.52 With very few exceptions, this situation prevails throughout the commentaries. This consensus speaks for itself: there is every reason to think that we are dealing with a genitive form. First of all, however, it is necessary to examine in detail the exceptional instances in the Nirukta where the extant text offers attributes in the genitive or the ablative case to analyses of the -teh and -eh types. Attributes to -teh/-eh forms in the genitive case In what follows I shall present all the material from the text of the Nirukta itself which may offer attributes in the genitive case to analyses of the -tehl-eh type. (1.17) athdpidam antarena padavibhdgo na vidyatel avasaya padvate rudra mrda/ itil padvad avasam gdvah pathyadanam I avater gatyarthasydso ndmakaranahl tasmdn ndvagrhnanti I If one phrases this as the editor, Lakshman Sarup, does, I see only one way of translating it: Xl Moreover, without it [= nirvacanasdstra] the [correct] division of words is not possible. 'Be gracious to the footed provision, O Rudra.' Here padvad avasam, [meaning] cows, [is] food for the journey, [and] asa is a noun-maker (i.e., suffix) of avatih which has the sense of moving; therefore [the authors of the Padapdtha] do not divide [ava and sam]. This, as will soon be evident, is not the only possible way of interpretation. Yaska's intention in this paragraph is to make it clear that a correct division of words, as carried out in the Padapdtha, is impossible without the science he represents. As his example he takep the word avasaya, dative sg. of avasa- (nt.) as attested in RV 10.169.1 avasaya padvate and thus not to be split up. He follows this up by contrasting it with ava saya, absolutive of ava^lso and as such to be divided in the Padapdtha.
52

Nllakantha reveals his view relatively seldom because he normally retains the compounds in °karmanah for metrical reasons whereas Durga and Skanda-Mahesvara often replace these by °arthasya

The universe of Yaska

113

Sarup (1921:17) translates this passage very differently: T h e footed wanderer, i.e. cows, provision for the journey: (avasdya) is derived from (the root) av, meaning to go, with the suffix asa\ it is therefore not analysed (in the Padapdtha)' Sarup may be doing the right thing in translating avasa- as 'wanderer', thus trying to follow Yaska. Sarup may be wrong, but let us for a moment assume that he is right. If he is, then avateh is unquestionably a genitive since it occurs with the attribute gatyarthasya. Sarup nevertheless translates as he usually does, taking avateh to be an ablative and conceiving of the ablative in his special way. The fact that gatyarthasya is genitive does not seem to bother him at all. But obviously other scholars too took this fact lightly. That artha is used here instead of the usual karman triggered off the following remark from B. Liebich (1922:4, note 1): 'Dieses vereinzelte gatyarthasya fiir gatikarmanah ist offenbar als lapsus calami eines Abschreibers erst spater eingedrungen.'53 This is not so obvious to me as it is to Liebich since artha occurs in similar constructions at several other places in the Nirukta. It is, of course, possible that rather than being changed by a scribe, we are in these cases dealing with later interpolations from a time when only the °a/t/za-alternative would occur to a Sanskrit speaker, although, obviously, he would be familiar with the other alternative from the Nirukta if he were to interpolate in that text. Be this as it may, the question is first of all whether the passage is of any real interest for the present investigation. The genitive, as is clear from my tentative translation above, could also be construed with aso ndmakaranah, and the passage would then no longer contain an analysis in -teh. Durga (D 1:131,1) glosses ndmakaranah with the technical term pratyayah 'suffix' which probably would not so easily construe with a genitive, although such a usage is met with in Panini's technical language, for example in the context of rule A 6.4.1 angasya. However, there is every reason to assume that Durga's gloss is mechanical in that he employs the standard technical term for a suffix well known from grammar without paying too much heed to the genitive construcjjyon. Yaska makes use of the term ndmakaranah, clearly his name for a krt|tiffix, also at several other places in the Nirukta.54 The passage is not commented upon further by Durga, but Skanda|lahesvara remark as follows (SM 1:104,14): evam atrdvasdyeti na syater avapurvasya lyabantasya riipam, dhdtupasargayor lyapas cdtrdrthdsambhavdt I padvata ity anena sdmdnddhikaranydc caturthyantatvdvagates calkim tarhil avater gatyarthasyaso namakaranah, ndma knyate yena sabdasya sa ndmakaranah pratyaya ucyatel tasmad etad padam navagrhnanti padakdrdh, avagrahasydrthabhedapratipddandrthatvdt, tasya cehdbhdvdtl

sporadic gatyarthasya for gatikarmanah has obviously occurred only later as a slip of the pen of a scribe.' kakso gdhateh I ksa in namakaranah / (2 2), gaur iti prthivindmadheyam I. I gdter vau ^ndmakaranah/ (2 5; cf Sayana at RV 1158 2), ksiram ksaratehl ghaser vero ndmakarana (2.5), anur anu sthaviydmsam I upasargo luptandmakaranah I yathd sampratil (6 22), m nau kasmdtl minotih srayatikarmd I thu iti ndmakaranah I thakdro vdl (7..29), seva i sukhandmal sisyatehl vakdro ndmakaranah I (10 17). These will be discussed below.

114

Indian semantic analysis

So, in this case avasdya is not a form of syatih preceded by the preverb ava and ending in [the absolutive suffix] LyaP (-yd), since the sense of root and preverb and LyaP is impossible here because of the syntactic agreement with 'padvate' and the understanding that it ends in a dative case suffix. How is it then? [The answer is] avater gatyarthasydso ndmakaranah', that by means of which a noun is made,55 that is called the ndmakaranah, suffix, of a linguistic item. Therefore th6y', namely the authors of the Pada[-pdtha], 'do not divide' this word, since a division has as its purpose to convey a division in meaning and that does not exist here. Here Skanda-Mahesvara seem to think that the genitive avater gatyarthasya is an expression similar to avasdyeti na syater avapurvasya lyabantasya rupam and that the words dso ndmakaranah are a separate addition to the real analysis since they only explain further the term 'ndmakaranah'. However, a complication arises with sabdasya which appears to be more or less superfluous if the sentence is phrased by placing a comma after yena. One could imagine that it reflects the genitive avater gatyarthasya, but one would then expect dhdtoh and not sabdasya. The interpretation offered is therefore not absolutely certain. If dso ndmakaranah really is to be conceived of as a syntactically separate addition, then avater gatyarthasya would be a completely ordinary analysis in -teh and unambiguously expressed in the genitive case. Nllakantha seems to think this is the case, presenting the analysis in the following Sloka (NSV 1.6.28): gatyarthasydvater dhur avasam pratyaye 'py ase56/ avanty anena gacchanti tatas sydd avasam hi tat II He says avasam is of avatih in the sense of gati ' motion' when there is also57 the suffix asa: avanti, that is, they move by means of it, so therefore this would be avasam. As stated already, I am not sure that this is how the Nirukta passage is to be understood, but the only way to achieve a clearer picture of the situation is by comparing this passage with the other instances where the term ndmakarana is used in the Nirukta. The following instances occur: * (1) Nir 2.2: kakso gdhatehl ksa iti ndmakaranah. This seems to support Nllakantha's interpretation, kakso gdhateh being the nirvacana proper followed by the additional information that 'ksa' is a ndmakarana: 'kaksali "armpit" is of/from gdhatih {4gdh "stir"); ksa is a/the ndmakarana'. Durga remarks (D 1:169,8-10): kakso gahater vilodandrthasya I kaksayor eva hi balena vilodayati stn dadhyddi dravyaml atra punah ksa iti namakaranah/ sarvam anyad ddyantaviparyayddi yathopadarsitam yathdsambhavam yojyam tatra tatra I 'kaksah is of/from gdhatih' which has the sense of stirring, for by means of thd strength of the very armpits a woman stirs such things as curd; but here 'ksa is | ndmakarana'. Everything else starting with the inversion of initial and final shoul<|
55 57

An analysis as karanasadhana. 56 Emended by the editor from °apasampratyatepy asaifi Nllakantha is not too careful with particles, evident also from the hi in pada d. *

The universe of Yaska

115

be construed wherever necessity arises as far as possible in the way that has been exemplified. It is clear from this that Durga understands ksa iti ndmakaranah as a separate addition. It is not clear to me, however, what he refers to by his additional remark, although I do not think it has any bearing on the way he understands the analysis. The term ddyantaviparyayah is well known from Nirukta 2.1 where Durga comments by listing several examples of metathesis, for example, scyutir ksarane tasyddyantaviparyayena stokah, '[the root] scyut when [the meaning] ksarana [is to be denoted]; [the form] stokah [appears] by inversion of the initial and final [sounds] of it' (D 1:161,1-2). It is therefore to be assumed that Durga refers to the phenomenon of metathesis, and the fact that this seems hard to apply directly to the example kaksah gdhateh is of less interest for the present purpose.58 What is important here is that kaksah is stated in immediate connection with gdhateh which therefore naturally is to be conceived of as "an ordinary -teh analysis and as such clearly taken as a genitive by Durga. There are no relevant remarks by Skanda-Mahesvara. (2) Nir 2.5: gaur iti prthivindmadheyaml yad duram gatd bhavatil yac cdsydm bhutdni gacchantil gdter vaukdro ndmakaranah. This passage is inconclusive. The first two analyses are as kartr59 'agent' and adhikarana ^'locus' in relation to the activity expressed by ^gam 'go; move': 'gauh is a name for the Earth, since it stretches out (lit. "is gone") far, and since living Beings move on it'. But the alternative analysis brings up similar problems as Before. Is okdro60 ndmakaranah to be construed with a genitive gdteh, or is it fan additional remark and gdteh a separate analysis of the ordinary kind? f Turning to the commentators, we do not find much to guide us. Durga remarks as follows (D 1:179,7): gdter vd 'gdn gatau' ity asya dhdtor okdro ridmakaranah pratyayah. This could easily comprise both alternatives: 'Or {gauh] is of/from gdtih (^gd), [and] "the syllable o is a noun-maker", [or] |uffix, of this root [which is listed Dhp 1.998:] "gdn when [the sense of] motion |is to be denoted]".' It is equally possible that the remark is to be read as two |eparate statements or, indeed, that it is to be read as one long statement where §ateh is to be construed with okdro ndmakaranah, the rest to be conceived of I I commentatorial additions structurally determined by the original syntax. Skanda-Mahesvara remark (SM 11:42,12): 'gdteh' vd stutyarthasya, %ykdrapratyayah. This is not very helpful, apart from making it clear that %oteh is to be understood as a genitive form.61 The phrasing suggested by the

^Possibly he thinks along the following lines: gd + ksa>kdgsa (by inversion)>kaksa. S So Durga (D 1:179,6) and Skanda-Mahesvara (SM 11:42,12). f H Or aukdrah, but see Durga below. ^Incidentally, gd is here given the sense of stuti 'praising', so Skanda-Mahesvara obviously Uiink of Dhp 3.25 gd stutau. They remark (SM 11:42,13-14): giyate stuyate 'sdv iti gauh, $dyanti vdsydm sthitd iti gauh I iha ca pilrvatra ca nirvacane knyaikatve kdrakabhed f giyate, that is to say, it is being praised, thus [it is called] gauh; or else: they praise who are located on it, therefore [it is called] gauh; and here, and in the preceding nirvacana, when there is [reference to] a single activity, there is a difference with regard to kdraka'.

116

Indian semantic analysis

>

editor, Lakshman Sarup, is peculiar. His inserting a comma seems to indicate that he thinks of the first part as a separate analysis and aukdrapratyayah as an addition. But as a compound the latter would not serve such a function. The question is therefore whether it is a compound or whether we are facing a mistake by the scribes or by the editor or, simply, a misprint. Why Sarup reads aukdra and not okdra like Durga does is also unclear. (3) Nir 2.5: ksiram ksaratehl ghaser vero ndmakaranah I mTram hi yathd. The first analysis is clear: 'ksTram "milk" is of/from ksaratih (^ksar "stream")'. The alternative analysis presents the same problems as before. Whether or not it is of any significance that the form ghaseh is based on the naked root and not on a finite form of the verb and therefore more attractive in a formal grammatical statement such as 'Tra is a noun-maker of ghasW is hard to tell. An explanation near at hand is that ^Ighas does not occur in the present. Moreover, there are several instances of forms in -eh instead of -teh in the Nirukta, and, as mentioned already, Katyayana sanctions both possibilities as valid ways of referring to a verbal root (A 3.3.108, vt. 2: ikstipau dhdtunirdese). There is accordingly every reason to treat the present instance as completely parallel to the preceding ones. The parallelism extends to the ambiguity of the expression; one could equally well translate 'or else it is of/from ghasih; Tra is a noun-maker, just as in usTram'62 and 'or else Tra is a noun-maker of ghasih, just as in usTram'. But there is one crucial difference in this passage. It is natural to see ksiram ksarateh and ghaser vd as parallel constructions and therefore to be interpreted in the same manner. Accordingly, we would have to divide the phrase differently from Sarup: ksTram ksarater ghaser vd/ Tro ndmakarana usTram iti yathd, 'ksTram is of/from ksaratih, or else it is of/from ghasih [and] Tra is a noun-maker, just as in usTram\ As is evident, it is even hard not to make the two possibilities overlap. Durga comments as follows (D 1:180,8-11): ghaser vero namakaranah63 / ader ghasl ddesah knyate atas tenajva siddharupena nirdisyate I dhdtvantaram iti vd ke citl jndpakam ca daddtil uslr^m iti yatha/ vasa kdntau tasya krtasamprasdranasya TrapratyayenosTram iti bhavatil tad dhi saugandhydt kdntam bhavatil 'ghaser vero ndmakaranah'''. [the root] ghas is made the replacement of ad 'eat',64 therefore [the analysis] is given by means of just this established form. [The meaning is:] or else, according to others, it is a different root, and he gives an indicator [for this possibility:] 'just as in usiram\ '[The root] vas [occurs] when beauty [is to be denoted]' (Dhp 2.70), [and] 'usTram' is of this, which has been subject to samprasdrana, through the suffix Tra; for this [usTram] is beautiful because of its fragrance.
62 63 64

A particularly fragrant plant-root. Rajavade (1921*145) gives the text as follows 'ghaser vd/ Tro ndmakaranah' - pratyayah Note his phrasing. A 2.4.37 lunsanor ghasl teaches that the verbal root ad is replaced by ghas before the drdhadhdtuka la-kdra replacements of the aonst and the desiderative; A 2 3 39 bahulam chandasi teaches that in the domain of Chandas this happens diversely

The universe of Yaska

117

The sentence vasa kdntau tasya krtasamprasdranasya TrapratyayenosTram iti bhavati, may give a hint that Durga considers Tro namakaranah syntactically independent of the genitive ghaseh in the Nirukta, but I am far from certain about this. The remarks of Skanda-Mahesvara are brief but illuminating (SM 11:45,3): ghaser vd, Tro namakaranah I adyate hi tat I irah pratyayah. Here it seems that Tro namakaranah is considered a separate addition. Skanda-Mahesvara first give a karmasddhana, an analysis as object: 'for this is eaten (adyate = *ghasyate65)\ Thereafter the commentary simply establishes that Tra is a suffix. This Nirukta passage itself, then, could point towards the conclusion that Tro namakaranah is a syntactically independent addition preceded by a -tehl-eh analysis. Skanda-Mahesvara certainly seem to hold such a view, maybe also Durga. (4) Nir 6.22: anuranu sthavTydmsaml upasargo luptandmakaranah, 'anuh "[something] minute; small", inferior to that which is abundant, [i.e., small], [is the] preposition [anu] with the suffix elided'. Clearly, this has nothing to do with the construction under investigation. (5) Nir 7.29: mithunau kasmdt I minotih srayatikarmd I thu iti namakaranah I thakdro vd, 'why mithunau "couple"? [Because it is of/from] minotih (Vmz "fix") which denotes the activity of depending; thu is the ndmakarana, or else it is the syllable tha\ This is quite revealing. In as much as there is no genitive construction here at all, thu iti namakaranah must be a separate addition. That the analysis is expressed in the nominative case is, however, atypical. (6) Nir 10.17: seva iti sukhandmal sisyateh I vakdro namakaranah I antasthdntaropalingiI vibhdsitagunah. This seems to support the view that we are dealing with two separate statements, sisyateh being the nirvacana proper followed by the additional information that the syllable va is a ndmakarana'. '[the word] sevah is a name for happiness [and] is of/from sisyatih (^sis "stir"); the syllable va is a/the ndmakarana, [and this suffix] occurs as a substitute66 in the place (antara) of the final sound (antastha) [s]; [and] optionally it takes guna\ Durga comments as follows (D 11:981,12-13): seva iti sukhanama/ tat kasmdt/ sisyateh/ vakaro namakaranah/ sisir dhdtuh, vakdrah pratyayah sisa ity asya. On the one hand, since Durga states tat kasmdt, 'why is that?', sisyateh must be the answer to that question and is accordingly not to be construed with the following. On the other, it is theoretically possible that he wants it both ways since vakdrah pratyayah sisa ity asya could be interpreted 'the syllable va is a suffix of this [root] sisa (ssis)', but sisa ity asya could certainly also be a rephrasing of the analysis sisyateh. * Skanda-Mahesvara do not add anything clarifying here, but it is significant that elsewhere in the commentary they occasionally use the term ndmakarana in constructions similar to the ones that have been examined. I shall adduce
65 66

See the remarks of Durga above. The term upahngin occurs only here in the Nirukta, and I am not confident about my translation

118

Indian semantic analysis

two examples. Commenting on Nir 3.11 tatra kutsa ity etat krntateh, 'among these [names for vajra] this kutsdh is of/from krntatih (Vfc/t "put")', they remark (SM 11:152,17): tatra kutsa ity etad yadd vajrah, tadd krntateh, chinattiti kutsah, rkdrasyokdrah so ndmakaranah, 'tatra kutsa ity etat: when it is a Vajra, then: krntateh, [that is to say,] it cuts, hence kutsah; the syllable u [occurs] in the place of the syllable r\ sa is a ndmakarana\ Similarly, at Nir 3.18 svd suydyil savater vd sydd gatikarmanah, 'svd "dog": it moves (ydyt) swiftly (su)\ or else it could be of/from savatih (^sav) which denotes the activity of moving',67 they comment on savateh (SM 11:183,10): gatyarthasyopadhdlopah ankdro ndmakaranah, '[or it is of/from savatih] which has the sense of moving; there is elision of the penultimate; the syllable an is a ndmakarana\ Both of these instances speak for themselves and may indicate that this is the way Skanda-Mahesvara conceive of the matter in general. To sum up, then, this little detour leaves us with the following information. In addition to Nir 1.17, there are five other relevant passages in the Nirukta. As for Nir 1.17, Nllakantha considers it an ordinary analysis in -teh. Nir 2.2 kakso gdhatehl ksa iti ndmakaranah appears to contain two separate statements, gdhateh being a -teh analysis. Thi$ seems to be supported by Durga. Nir 2.5 gaur iti prthivindmadheyam . . . gdter vaukdro ndmakaranah is inconclusive, also in the commentaries. Nir 2.5 ksirarn ksaratehl ghaser vero ndmakaranah seems to indicate that we are dealing with an analysis in -teh, taken to be a genitive by Skanda-Mahesvara. Nir 7.29: mithunau kasmdtl minotih srayatikarmd I ihu iti ndmakaranah I thakdro vd contains no possible genitive construction, but is quite revealing in that minotih srayatikarmd and thu iti ndmakaranah must be syntactically separate expressions. Nir 10.17 seva iti sukhandma I sisyateh I vakdro ndmakaranah etc. also seems to indicate that we are dealing with two separate expressions. Finally, such a view is supported also by usage in the commentary of Skanda-Mahesvara. On this Evidence one ought to conclude that Nir 1.17 padvad avasam gdvah pathyaddnaml avater gatyarthasydso ndmakaranah also consists of two separate statements, avater gatyarthasya being an analysis of the -teh type. Still, I cannot help feeling that one would have to twist the construction quite hard if one were to achieve this result. Note, however, that if we really are dealing with an analysis of the -teh type, we are here dealing with a genitive case suffix. (3.18) simhah sahandtl himser vd sydd vipantasyal sampurvasya vd hantehl samhdya hantiti vd/ simhah 'lion' [is so called] because of the conquering (sahandt: sahm); or it could be of himsih (^hims 'injure') which has been inverted; or else it is of hantih (V/ian 'kill') preceded by [the preverb] sam\ or [it is so called because] leaping up (sam^lha) it kills (^Ihan).
67 68

Nigh 2.14 lists savati under gatikarmanah The meaning of sah is uncertain. Dhp 1 905 gives saha marsane and 4 20/1 saha suha cakyarthe The commentators explain its meaning as abhibhavah, 'be powerful, conquer'

The universe of Ydska

119

This passage contains one analysis of the -andt type and two analyses of the -tehl-eh type, both of these with attributes in the genitive. The short form himsih may have been resorted to because hinastih would not so readily create associations to hims. It is, however, possible to argue that this analysis is an interpolation. Durga does not quote or comment upon himser vd sydd vipantasya. The source for its later addition could be Patanjali's Mahdbhdsya (Mbh 11:87,16) where the analysis himseh simhah occurs. The assumption that the analysis is an interpolation is supported also by Durga's remark (D 1:318,13): simhavydghrasabdayos tu anulomam eva nirvacanam0', 'the analysis of the words simhah and vydghrah goes only with the grain (anulomam, that is to say, is not inverted)'. This indicates that Durga knew of the analysis himseh but not from the Nirukta. It seems reasonable to think of himseh as a more formal grammatical analysis, necessitating a translation 'simhah is [a form] of himsih which has been inverted'. Also in the case of sampurvasya vd hanteh/ samhdya hantlti vd there is ample reason to doubt the text since these two phrases cannot represent more than one analysis in as much as both parts refer to sanAhan. If samhdya were a separate analysis (sanvlha) then the subsequent hand is slightly problematic. As far as I can see, there are three possible solutions to the problem: (1) sampurvasya vd hanteh is a later addition, inserted to give a formally more precise explanation of samhdya hantlti vd which would thus represent the original text; or (2) iti vd or at least the last vd is a later addition; or (3) sampurvasya va hantehl samhdya hantlti vd provides us with two possible analyses - in the first, sam is the upasarga of V/ian, in the second, sam is the upasarga of V M , but the following hanti is what provides the second syllable of simha. Durga remarks (D 1:318,5-6): sampurvasya vd hanteh upasargasyetvena I samhaya hantlti va vaiydkarandndm esd vyutpattih, 'or [simhah] is of hantih preceded by [the preverb] sam, in as much as [the sound] / [appears instead] of [a in] the preverb: samhdya hantlti vd\ this analysis belongs to the grammarians'. The editor of the text, H.M. Bhadkamkar, does not indicate sampurvasya vd hanteh typographically as a pratika which would be his usual practice. Whether or not it is a coincidence or he in fact considers the words to be Durga's own, is hard to tell. It is certainly possible that the words sampurvasya vd hanteh have crept into the extant text of the Nirukta on the basis of Durga's commentary. The other possibility, that vd or iti vd is a corruption, is expressed by Rajavade (1940, notes p. 508): 'iti vd after hanti seems interpolated both in the ?Jext and in Durga's Vrtti'. Skanda-Mahesvara do not read a vd after hanti. On ihe contrary, we meet with the following remark (SM 11:183,14): °hantlti piirvasyaivdrthavacanam I ato vety apapdthah,' [samhdya] hanti is just a statement explaining the meaning of the preceding [sampurvasya vd hanteh]; a vd after that is a corrupt reading'. In any case, the passage contains no more than two analyses: (1) sahandt and (2) sampurvasya vd hanteh/ samhdya hanti (iti) or only samhdya hantlti

120

Indian semantic analysis

vd. If sampurvasya vd hanteh is Yaska's own analysis, we are dealing with a formal analysis in -teh clearly expressed in the genitive case. This is clearly how Skanda-Mahesvara understand it, with samhdya hanti added for further elucidation. (4.10) saktuh69 sacateh/ durdhdvo bhavatil kasater vd sydd vipantasya I vikasito bhavatil saktuh 'coarsely ground barley5 is of/from sacatih (4sac 'accompany; stick together'); it is difficult to rinse; or it could be of kasatih (^kas 'move; split'70) which has been inverted; it is split open. Here there is no option at all; vipantasya is an attribute to kasateh which accordingly has to be interpreted as^a genitive. The reading is confirmed by the commentators who add nothing further of interest (D 1:372,6; SM 11:221,1-2). Unfortunately, this is not all there is to this seemingly clearcut^case. In the Mahdbhdsya the passage is quoted almost verbatim, but with-an ablative viparltdt (Mbh 1:4,12): saktuh sacater durdhdvo bhavatiI kasater vd vipantdd vikasito bhavati. Matters are not made easier by Kaiyata's remark (MbhP I:46a): sacater itil 'saca secane1 ity asya, 'sacateh: [that is,] of [the verbal root] saca (sac) [which according to Dhp 1.176 occurs] when secanam "sprinkling" [is to be denoted]'. This makes it clear that he considers sacateh to be a genitive. He goes on to say (ibid.) kasater itiIprsodardditvdd varnavyatyayah, 'kasateh: there is transposition of sounds because [saktuh] belongs to the prsodarddi group [i.e., gana on A 6.3.109]'. Since Kaiyata takes sacateh to be a genitive it is reasonable to assume that he would also take kasateh as a genitive. It is noteworthy that he does not cite the word vipantdt. Nagesa's Uddyota does not contain anything indicative of his view. It is hard to see any clear way out of this mess. As matters stand, however, the extant text of the Nirukta and the commentators of that text clearly read vipantasya, and kasateh is accordingly to be considered a genitive form. (6.30) vikato11 vikrdntagatir ity aupamanyavah I kutater vd sydt [vipantasya] vikutito bhavatil vikatah 'horrible [famine]' [is so called] because its ways of moving are crooked, according to Aupamanyava; or else it could be of kutatih i^lkut 'be crooked') which has been inverted; it is very72 crooked. Sarup gives vipantasya in brackets as it is omitted by several MSS. Rajavade (1940:97) notes it as a variant reading. Bhadkamkar (1918=D 1:697,2) includes it in the text with the following note: 'vipantasya has not
69 70 71 72

The context is RV 10.71.2. The meaning is unclear. The verb kasati is listed under gatikarmanah Nigh 2 f4 and Dhp 1.913 lists kasa gatau. A more explicit meaning than 'moving' is required in this context, though. The context is RV 10.155 1. The exact meaning of vikutitah is not clear Durga glosses it (D I 699,10) by kubjibhutah 'has become crooked/humpbacked'

The universe of Yaska

121

been noticed by Durga arid is omitted by three Mss. of Bi [= Bibliotheca Indica edition]. The omission appears to be preferable. We retain it, however, as it is supported by an overwhelming majority of Mss.' Bhadkamkar is certainly right in claiming that the omission of vipantasya would be preferable. For though several MSS read it, it gives no sense in as much as an inverted kut would give you tuk and thus not get you any nearer to vikatah. SkandaMahesvara (SM 11:494,3), however, rather surprisingly claim the inversion refers to a occurring in the place of u. They do not cite the genitive form vipantasya, though. Skold (1926:322) remarks: 'No inversion is necessary to explain the etymology; the grammar wants vipantdd; so vipantasya should be cancelled.' The grammar does not want anything in particular, and his remark shows how narrow-mindedly scholars have thought an ablative is the only conceivable way to interpret a form in -teh. Once again, we are in an inconclusive situation. If we believe vipantasya is a genuine reading, there is no doubt about kutateh being a genitive form, if not, then there is no evidence for this being so. (9.11) ratho ramhater gatikarmanah I sthirater vd sydd vipantasya I rathah 'chariot' is of/from ramhatih i^ramh 'hasten') which denotes the activity of moving; or else it could be of sthiratih {^sthir) which has been inverted. Concerning sthiratih, there is no such verb in the Nighantu, nor is a corresponding root listed in any dhdtupdtha. Skold (1926:309) nevertheless indicates sthir as the root and gives the meaning 'to stand firmly'. S. Varma (1953:110) indicates that this is the root, but he gives no meaning for it. Sarup ,(1921:142) translates: 'Rathah (chariot) is derived from (the root) ramh, rineaning to speed, or from sthira by metathesis', with the note: 'i.e. sthira> Hhara, and by metathesis ratha? Skanda-Mahesvara confirm the reading (SM \ III: 153,12): sthirater vd viparltdksarasya, that is 'with inverted syllables'. Purga quotes the Nirukta text but does not add anything of interest. Here, then, we have an attribute in the genitive case without anything in particular to cast doubt on it. (10.10) parjanyas trpehl ddyantaviparltasyal

[The name of the deity] parjanyah is of trpih (^trp 'be satisfied') which has the first and the last [sound] inverted. Here we have a clearcut case of a genitive, although the analysis, expressed by a short form of the root, seems more than usually designed to bring out the elements involved in the grammatical derivation of the word parjanyah. This claim is supported by the remarks of Durga (D 11:970,9-10): parjanyas trper adyantaviparltasya/ 'trpa trptau1 ity asya dhdtohl sa katham bhavatil ddy~ !antaviparyayena, '"parjanyah is of trpih (^Itrp 'be satisfied') which has the first and the last [sound] inverted", [that is, it is] of this root [which is listed (Dhp 6.24 and 10.276:] "trp when [the sense of] satisfaction [is to be denoted]".

122

Indian semantic analysis

How does this [derivative] come about? Through inversion of the first and the last [sounds]'. Skanda-Mahesvara, confirming the genitive reading, offer a similar remark (SMIV: 13,15). Sayana quotes the analysis at RV 5.83.1. (10.31) madhu dhamater vipantasyal madhu 'honey; sweetness; water (?y73 is of dhamatih {^dham 'blow; melt') which has been inverted. The context is RV 4.38.10. The commentators come up with various explanations of the semantic relationship between madhu and Adham which need not concern us here. They do, however, confirm the genitive reading (D 11:1008,11; SM IV:38,14), so dhamateh is indisputably a genitive form. (11.5) cam rucer vipantasyal cdru 'bright; lovely'74 is of rucih {^iruc 'shine') which has been inverted. The genitive viparltasya is confirmed by Durga (D 11:1040,4). SkandaMahesvara (SM IV:57,8) prefer to rewrite by viparyayena 'through inversion'. Skold (1926:246) remarks: 'Gen. for abl. We expect rocater' Why take this for granted? Moreover, there is no reason to expect rocateh; ruceh.xs perfectly acceptable and, indeed, the form cameh occurs in the previous sentence. This, then, is another instance of an analysis expressed with a genitive case ending. (12.26) visam ity udakandmal visndteh I [vipurvasya sndteh suddhyarthasaya I] vipurvasya vd sacatehl dydvdprthivyau ca dhdrayatil 'visani' is a word for water,75 [and] is of visndtih (vi^lsnd 'bathe'), [that is to say,] of sndtih preceded by [the preverb] vi with the sense of purifying; or else it is ofsacatih (^Isac 'accompany') preceded by [the preverb] vi; it supports heaven and earth. The reading vipurvasya sndteh suddhyarthasaya is omitted by several MSS and therefore bracketed by the editor. It is, however, met with in the commentary of Durga (D 11:1132,12-13). Skanda-Mahesvara, on the other hand, do not support the reading. Instead we find (SM IV: 125/7): visndteh saucdrthasya, 'of visndtih which has the sense of purification'/This clearly suggests an interpretation of visndteh as a genitive form, although the addition saucdrthasya seems to be their own elucidation. It is therefoie possible that the words vipurvasya sndteh suddhyarthasaya have been included in the text of the Nirukta on the basis of Durga's commentary. The second analysis, vipurvasya vd sacateh, is supported by both commentaries (D 11:1132,13; SM IV: 125,7). On the basis of the Nirukta itself, then, we have a clear case of an analysis of
73 74 75

madhu is listed under udakandmdni at Nigh 1.12. * • Yaska quotes RV 10 85 19 as a stanza pertaining to it, but it is not clear who or what is referred^ to in this context So Nigh 1.12, so also Grassmann ('Wasser, Flussigkeit') but rejected by Mayrrjofer who thinks the only meaning of this word is 'poison'.

The universe of Yaska

123

the -teh type clearly to be understood as a genitive form, and maybe one more instance of the same. Sarup (1921:192) nevertheless translates: The word visam is a synonym of water, derived from the verb vi-snd from snd preceded by vi, meaning to purify, or from sac (to accompany) preceded by vi.' This concludes the material from the Nirukta offering possible attributes in the genitive case to technical analyses of the -tehl-eh type. So where do we stand? We have eleven instances of possible attributes. Of these I would dismiss himser vd sydd vipantasya (3.18) and vipurvasya sndteh suddhyarthasaya (12.26), maybe also kutateh . . . vipantasya (6.30), as later additions. Moreover, avater gatyarthasydso ndmakaranah (1.17) I would prefer to keep aside as a special case, although the rest of the material from the Nirukta and the remarks of the commentators point to its inclusion among genuine cases. This leaves us with eight or nine seemingly clearcut cases of attributes with a genitive case ending, all of them supported by the commentaries and none of them contradicted by evidence of any kind with the exception of kasateh . . . vipantasya (4.10) which the Mahdbhdsya cites (Mbh 1:4,12) with the ablative vipantdt, although Kaiyata seems to conceive of it as a genitive. The important thing is that these eight or nine cases do occur. It is of course always possible to argue that they are the result of later tampering with the text; it is impossible to guard against interpolations. But even in that case the fact remains: why was the genitive case chosen? It is now necessary to investigate whether the Nirukta itself offers contrary evidence in the form of attributes with an ablative case ending. Attributes to -teh/-eh forms in the ablative case (3.13) mahanndmdny uttardni pahcavimsatih I. . . tatra vavaksitha vivaksasa ity ete vakter vd vahater vd sdbhydsdtl This is a difficult passage. Nigh 3.3 lists the forms vavaksitha and vivaksase under mahanndmdni, names for 'great'. This is rather surprising since they are [verbal forms. The form vavaksitha I would be inclined to take as the 2. sg. perfect pf^lvaks 'increase', however Yaska relates it to Vvac 'speak'. The second form, Mvaksase, is even worse. Grassmann notes: "vivaksase lasst sehr verschiedene Deutungen zu [von vaks mit vi, Desid. von vac oder vah, 2. s. med. oder Dat. AQS Inf.; der Sinn nicht klar, da es nur als Refrain vorkommt] \ 76 Geldner remarks at RV 10.21.1: 'vi- vivaksase ist das Desider. von vi-vac mit der ratselhaften Endung se, die hier einmal an den Desiderativstamm antritt.77 Vgl. das gleichtfalls vereinzelte Intens. cdkrse 10,22,1. Naigh. 3,3 zahlt vivaksase unter den P6 'vivaksase allows for very different interpretations [from vaks with vi, desid of vac or vah, 2 sing, middle or dat of the inf., the sense is not clear because it only occurs as a refrain].' '7 Geldner adds in a note that the accent is explained by the separation of the refrain by the « inserted pdda d ,

124

Indian semantic analysis

Wortern ftir "groB" auf, Yaska 3,13 sieht darin eine reduplizierte Form von vac oderva/i.' 78 In addition to this we face the possibility that sabhyasat could be taken as an adjective simultaneously qualifying both vakteh and vahateh: These one would accordingly have to interpret as ablative forms. The question remains, though, whether this is a necessary and correct interpretation. The word sdhhydsa occurs only once in the Nirukta. However, several times Yaska uses the adjective abhyasta 'reduplicated'. This refers directly to the nominalised verb he makes use of in the analysis.79 Moreover, we meet with the noun abhydsa 'reduplication [in a verbal stem]; reduplicated syllable'.80 If all Yaska wanted to do was to qualify further vakteh and vahateh, he could simply have used the adjective abhyasta. On the other hand, there is reason to assume that he wanted to make clear that vavaksitha and vivaksase are reduplicated forms and that this is the reason they are to be related to vaktih or vahatih. In that case sdbhydsdt would refer to vavaksitha and vivaksase directly, that is to say: 'because [these are forms] with reduplication'. Support for this interpretation is given by I^enou (1942:377): 'sdbhydsa forme verbale "redoublee" N. Ill 13 (ex. vivaksase)'}1 However, one could also translate: 'vavaksitha and vivaksase are from a reduplicated form of vaktih or vahatih\ although in that case one gets even more severe difficulties with regard to number. Sarup interprets the passage in the following way (1921:46): 'With reference to these, the two words vavaksitha and vivaksase are the reduplicated forms either of (the root) vac (to speak) or of vah (to carry).' It is noteworthy that here, where he really has every opportunity to stick to his usual ablative interpretation in the sense 'derived from [the root]', he translates as if we were facing a genitive form. Durga comments as follows (D 1:292,6-7): tatra tasmin mahanndmagane vavaksitha vivaksase ity ete dkhydtel te punar ubhe vakter va vahater va dhdtoh sabhyasat, 'moreover, both of these are from vaktih or vahatih, [that is to say,] from a root with reduplication'. It is, however, possible to translate: ' "with regard to these", [that is,] with regard to the group of names for "great", "these two" [forms] "vavaksitha [and] vivaksase" are verbs; moreover, both of these "are [forms] of vaktih or vahatih because" the root "has reduplication"'. The passage is therefore not very illuminating. Turning to the commentary of Skanda-Mahesvara, we meet with the following passage (SM 11:156,11-13): tesu vavaksitha vivaksase ity ete dkhydtapade iti sesahl te tu vakter va vahater va sabhyasat - sanndbhydsasya dve api na yathdsamkhyena I artharupasya cobhaya78

81

*vi — vivaksase is the desiderative of vi-vac with the enigmatic ending se which attaches itself to the desiderative stem. Compare the also sporadic intensive cdkrse 10 221 Nigh 3 3 lists vivaksase among the words for "gxt2X">,{mahatah), Yaska sees it as a reduplicated form of va 79 80 or vah ' See, for example, Nir 2.12, 4.23,' 4.25 For example, Nir 2 2, 2 3. 'sdbhydsa "reduplicated" verbal form Nir 3 13 (ex. vivaksase)'

The universe ofYaska

125

tra sambhavat, atisayavato vacandd vahandd vd mahattvasydpi pratiteh, mahanndmasupapannah pdthah I No interesting variant readings occur. In a note to sdbhydsdt, the editor remarks: agre *sanndbhydsasya' ityasyasthdne' sdbhydsasya' itipdthahyuktahl sdbhydsdd iti sasthyarthepancamiiti bhdvah. This amounts to saying: 'In what follows (= with regard to the following word), the correct reading is "sdbhydsasya" instead of this "sanndbhydsasyd". This means that [according to the commentary] the fifth case ending of sdbhydsdt occurs in the sense of a sixth.' It is not quite clear to me whether Sarup means to say that it is the view of Skanda-Mahesvara that the ablative occurs in the sense of a genitive, or whether he himself thinks so, uncharacteristic though this would be. As seen above, he certainly translates vakteh and vahateh, probably also sdbhydsdt, as if they were genitive forms. Sarup is probably right in considering sanndbhydsasya a corrupt reading, and sdbhydsasya is indeed a good candidate. Another, less likely, possibility would be to read sannabhydsasya, san being the technical term for the desiderative suffix. Reading sdbhydsasya, one could translate: 'With regard to these, the two [forms] vavaksitha [and] vivaksase' are verbs, thus [the statement] is to be completed. As for these two, [they] 'are from/of vaktih or vahatih because they have reduplication/which have reduplication', [that is to say, they are forms] of [these two roots] which have reduplication, although the two [words] are not [derived] respectively [from those two roots]. And since meaning and form [?] are possible in both cases, [and] since there is a clear understanding of 'greatness' because of excessive talking (vacanam) and carrying (vahanam), [their] reading among the names for 'great' is acceptable. Skanda-Mahesvara continue by quoting examples intended to illustrate the occurrence of vavaksatha and vivaksase in the sense of 'great'. It is clear from this passage that sdbhydsdt is glossed by a genitive form. Whether this gives us reason to doubt the reading sdbhydsdt with an ablative ending is hard to tell. It may be of no importance, but I do find it worthy of notice that the analysis is immediately followed by an analysis undoubtedly expressed with a genitive case ending: grhdh kasmdtl grhnantiti satdm. While the reading in the extant manuscripts of the Nirukta is undisputably an ablative sdbhydsdt, I am nevertheless in doubt as to its interpretation and therefore reluctant to consider this a clear-cut case. (3.20) ardham harater vipantdtl dhdrayater vd sydtl uddhrtam bhavatil \ rdhnoter vd sydtl rddhatamo vibhdgahI At first sight vipantdt appears to be an adjective related to harateh which accordingly would be an ablative form. Renou (1942:494) remarks: 'vipanta "metathese" de phonemes N. passim; "phenomene inverse" R. XIV 42 (800), | 6 (804), Sarvanukram., Ping, passim'. Abhyankar (1961/1977:356) has the following entry: 'vipanta (1) in the opposite or reverse way; cf. viparitdc ceti vaktavyamlpdrdvdnnah M.Bh. on P. IV.2.93, Vart. 2'.

126

Indian semantic analysis

Surprisingly, Rajavade (1940, notes p. 525) remarks: 'ardha comes from hr {harateh). [. . .] harateh vipantasya [sic] = of hr the two parts of which are reversed'. Clearly, Rajavade considers vipanta an adjective, but he quotes the text with the genitive form vipantasya. This is all the more surprising since he reads vipantat in his edition of the text (ibid :53). Could he have changed it to make it translate as an adjective? Of course, most past participles can be used as either. If it is conceived of as an adjective, vipantat ought to refer to dhdrayateh as well since the same phonetic complication occurs in this analysis, as noted by Varma (1953:73): 'arddha- (°arddham dharayater vd sydd uddhrtam bhavati, N. Ill, 20) is traced to Adharay with metathesis, lit. "that which is carried away".' But although vipantat is neither repeated nor stated in the dual, it may have been intended to work with both analyses. This would work more easily if it is conceived of as a substantive. It is, of course, also possible that dhdrayateh is a later addition, or that something is wrong with the text here. Sarup (1921:53) interprets: 'Ardha (half) is derived (1) from (the root) hr (to take away) by metathesis.' This gives no clear indication as to whether he interprets vipanta as an adjective or a substantive. It seems possible to interpret as a substantive, that is to say, vipantat 'because of inversion'. The forms in -teh could then be genitive as well as ablative, and the whole passage could be interpreted: ardham 'half is of/from haratih {\lhr 'take'), since there is inversion; or it could be of/from dhdrayatilf2 (^Idhr 'carry', caus.): it is carried away/removed; or it could be of/from rdhnotih i^rdh 'thrive'): it is the richest part [into which something can be divided]. Unfortunately we do not have much material from the commentaries here. Skanda-Mahesvara are non-existent, and Durga is suspiciously brief. He gives the text as above, adding only (D 1:331,10) °uddhrtam hi tat bhavati, and gloss-, ing rddhatamah by abhisampannah. On the evidence available here I would hold that the passage is inconclu-; sive, but it is certainly possible that vipantat is an adjective and that haratetii therefore is an ablative form. This possibility may be strengthened by th% passage to be discussed next. $ f (4.25) amhatis cdmhas cdmhus ca hantehl nirudhopadhdt I vipantat I The context is RV 1.94.2 where amhatih 'distress' occurs. Here it is clea$ that nirudhopadhdt can only be a masculine bahuvnhi meaning 'with the!* penultimate put aside [i.e., to the front]', and vipantat would most naturally bej taken as an adjective as well. But this is not absolutely certain. The lack of am
82

: -h The analysis dharayateh is a bit uncertain since the text without a danda would 0 °vipantdd dharayater . It is therefore just as possible that we are dealing with the causative I hr, hdrayateh. MW traces uddhrta to ud^lhr. One of the meanings attributed to the causativ hdrayati by MW is 'to have taken from one's self, be deprived of, lose'.

The universe of Yaska

127

expected ca 'and' could be explained by the very fact that the two expressions are to be understood differently, the first as an adjective and the second as a substantive giving the reason why the penultimate has been put aside. In any case hanteh has to be considered an ablative form. We could translate: amhatih, amhah, and amhuh [all meaning 'distress, trouble, anxiety'] are from hantih (^Ihan 'strike, kill') with the penultimate put first as a result of inversion/ and inverted. Apart from confirming the reading in a pratika, Durga has no remark of interest here. Skanda-Mahesvara, on the other hand, comment as follows (SM 11:270,18-20): hanter upadhd yd akdrah sanniskrsya*3 udhah prdpito nita ddau yasya sa nirudhopadhah I ydv api hakdranakdrdvayavau tdv api vipantau yasya sa viparltdvayavatvdd vipantahl The [root] han, of which the penultimate sound, a, is extracted and brought to the front, that is nirudhopadhah. And that of which those two parts which are the syllables ha and na have been inverted, that is [called] inverted because its parts have been inverted. This makes it quite clear that not only is nirudhopadhdt considered a bahuvnhi and thus an adjective, but vipantdt is as well. There is every reason, then, to take this passage as an instance of at least one and possibly two attributes in the ablative case. (5.26) tdlu taratehl tirnatamam angaml latater vd sydd [lambakarmanahf4 vipantdt I yathd talaml latety aviparyayahl tdlu 'palate' is of/from taratih (Vfr 'cross over'): it is the most crossed over organ [of the body]; or it could be of/from latatih (Vto 'be long'85) [which denotes the activity of hanging down] because of inversion/which has been inverted, just like talam 'surface'; laid 'creeper' is [formed] without inversion. Again, we face the same problem as in the two preceding passages, but here the use of the substantive viparyaya in forming a compound with the negative particle a may indicate that vipantdt is to be interpreted as an adjective agreeing with latateh which then would have to be an ablative form. Speaking against tjus is the inserted lambakarmanah which would make such a construction Cumbersome, and again one would miss a ca 'and' as in the preceding passage ||.25); however, the insertion is excluded by a number of MSS (see note). Durga IP 1:566,12) glosses vipantdt by ddyantaviparyayena 'by inversion of first and last'. This may indicate that he takes vipantdt as an ablative of cause, although, ^ The edition reads sanmkrsya, but this is clearly explaining the nih part of nirudhopadhah and we want the meaning 'drawn out', so I think this small emendation is necessary. m Omitted by several MSS; not read by Durga or Skanda-Mahesvara; included by Bhadkamkar 1918 (=D I); excluded by Rajavade 1940. **,No V/af is met with in the Dhatupathas or the Nighantu; Palsule (1955) lists it as a Sautra root.

128

Indian semantic analysis

as noted above (11.26), Skanda-Mahesvara chose to explain a genitive vipantasya by viparyayena (SM IV:57,8). But Durga adds as an explanation of latd (ibid., 13-14): asyeiva latater lambandrthasya latety etac chabdarupam aviparyayoS6 'viparyayena bhavati, '[also] of this very latatih, which has the sense of hanging down, is this linguistic form latd which is without inversion, [that is,] it is [formed] by non-inversion'. This statement suggests that Durga considers latateh a genitive form and vipantdt an ablative form meaning 'by inversion; as a result of inversion'. On the evidence available, I consider the present case inconclusive, although we might have an attribute in the ablative case here. (6.3) salalukam*1 samlubdham bhavatiIpdpakam sararukam^ vd sydtl sarter abhyastdft9/ iti nairuktdhl

salalukam means [something] confused; [it means something] bad according to the Nairuktas; or else it could be sararukam, [that is to say,] from sartih (Vsr 'flow') which is reduplicated. First of all, the meaning of salalukam, and that of sararukam as well, is unclear. Mayrhofer notes: 'salaluka-, nur RV 3, 30, 17 °am:unklar. Meist durch "Umherschweifen, umherschweifend" iibersetzt (PW, bzw. Grassmann); glaubhafter Pischel, Ved. Stud. Ill 204f. (s. auch Oldenberg, Noten I 239, Geldner I 365): "Nachsicht, Geduld". In diesem Falle ware vielleicht an sar"sich erstrecken" zu denken (s.u. sisarti)'90 The commentators have rather lengthy discussions of the meaning, but they need not concern us here. What is of immediate interest is that we seem to have an instance of an attribute in the ablative case here, although some caution is called for. In spite of discussing the passage at length, Skanda-Mahesvara neither quote nor discuss sarter abhyastdt. Durga remarks as follows (D 1:581,17-582,2): evam atra salalukam iti lubher mohandrthasya sabdasdrupydt samlubdham ity upapadyatel sarter apy atra sdrupyam astlti sararukam va syad ity uktam bhdsyakdrena I raksovisesanam caitat tasya hi sammoho vd ndsanqm vdbhistam itui * ,' Thus, in this case, because of similarity in sound with [the root] lubh which has the sense of confusion, 'salalukam' is established [with the sense of] 'samlubdham'. Also, there is similarity [in sound] with [the root] sr here, [and] thus it has been stated by the author of the Bhasya^that 'alternatively it could be sararukam'. And this is a qualifier of a demon who wishes confusion and destruction.
86 87 88 89 90

Excluded by several MSS and omitted in the edition of Rajavade (1921) The context is RV 3 30 17, salalukam occurs Nigh 4 3 So read for sararukam with several MSS, Durga and Skanda-Mahesvara. A fair number of MSS read ahyasndt 'salaluka-, only RV 3 30.17 °am : unclear Mostly translated as "aimless wandering, wandering aimlessly" (PW, Grassmann, respectively), more credible is Pischel, Vedische Studien III 204 f. (see also Oldenberg, Noten 1239, Geldner 1365) "forbearance, patience". In this case one could perhaps think of sar- "to stretch oneself (see under sisarti).'

The universe of Yaska

129

If the genitives lubher mohandrthasya and sarteh are constructed with sabdasdrupydt and sdrupyam, respectively, one would have to translate as above, and I do think that is the most likely interpretation. In that case we have no technically expressed analyses here. If the genitives are not constructed with sabdasdrupydt and sdrupyam, however, it is possible that Durga rewrites the analysis of salalukam into the more technical expression lubher mohanarthasya, clearly a genitive form, and takes sarteh as an alternative analysis.91 In any case Durga does not quote the word abhyastdt at all, and the analysis sarteh is attributed to the bhdsyakdra, the author of the Bhasya. In grammatical literature this is normally no other than Patanjali, the author of the Mahdbhdsya. This analysis is not found in the Mahdbhdsya. In fact, it is evident that Durga thinks of the Nirukta itself as the Bhasya in question, and that it is Yaska he has in mind. This is his common practice, for example at Nirukta 6.8 (D 1:615,10) and Nirukta 6.21 (D 1:667,16). The Nirukta passage is quoted in the form given above by Venkata Madhava and Sayana in their respective commentaries on the Rgveda. The fact that abhyastdt is not commented upon or quoted by any of the commentators warrants the conclusion that the last analysis is an interpolation in the extant text of the Nirukta. This conclusion is supported also by the fact that the identification of salalukam with pdpakam is attributed to the Nairuktas among whom one would tend to include Yaska. There is therefore no evidence that sarteh, even if genuine, would be an ablative form. That some later interpolator has added the ablative abhyastdt is, of course, of interest in itself, but (Joes not tell us anything about Yaska's own position. (7.9) suvidatram92 dhanam bhavatil vindater vaikopasargdtI daddter93 vd sydd dvyupasargdt I suvidatram means wealth; either it is from vindatih (Vv/d 'find') with one preverb, or it could be from daddtih (^dd 'give') with two preverbs. Here we seem to have two clear-cut instances of an analysis of the -teh type ^occurring with an attribute in the ablative case. Both readings are supported :by the commentators (D 1:772,7-9; SM 111:55,8). Some caution seems called for, though, since there are instances in the extant text of the Nirukta where compounds with the word upasarga as their final member seem to be later additions. An example is Nir 6.11: agham hantehl nirhrasitopasargahl dhantlti. Sarup (1921:97) translates: 'Agha (evil deed) is derived from (the root) han with the preposition a shortened, i.e. it kills.' That is to say, Sarup translates as if nirhrasitopasargah were a bahuvrlhi In syntactic agreement with hanteh, a possibility precluded by the fact that it In that case one would have to translate the passage 'Thus, in this case, because of similarity in sound, salalukam is of lubhih which has the sense of confusion, [and thus] samlubdham is [an] appropriate [synonym]. Also, it is of sartih [in as much as] there is similarity here, [and] thus it has been stated by the author of the Bhasya that alternatively it could be sararukam' 92 93 The context is RV 10.17.3 = AV 18.2.54. v.l. dadhatir0.
91

130

Indian semantic analysis

appears in the nominative case, admittedly itself a rather curious feature. Durga does not comment upon this analysis at all, a fact which indicates that we are dealing with an interpolation. And Skanda-Mahesvara remark (SM 11:425,14): agham hanteh purvasydno hrasvatvena, '"agham is of/from hantih", because of shortening of the preceding dN (= the preverb d)\ (7.23) dasyur94 dasyatehl ksaydrthdtl upadasyanty asmin rasdhl upaddsayati karmdnil dasyuh 'Dasyu' is from dasyatih (^das 'waste, exhaust') which has the sense of destruction: the juices are exhausted m him, or he causes [ritual] actions to be wasted.95 Here we seem to have a straightforward attribute in the ablative case, were it not for the fact that Durga does not seem to support the text at this point. He quotes the word dasyum with an accusative case ending which is how it appears in the stanza in question (RV 1.59.6), and goes on to say (D 1:812,5-6): ddsayitdram upaksapayitdram rasdndm susyanti hi tadanudgame sasydni96 karmandm vopaddsayitdram andvrstidvdrena tarn dasyum0, 'him who causes exhaustion, [that is to say,] who causes waste of the juices, for the crops dry up when he does not go away; or else, he is the extinguisher of [ritual or agricultural?] actions by causing the lack of rain - that Dasyu'. There is no reflection here of the wording ksaydrthdt as met with in the extant text of the Nirukta. The remarks of Skanda-Mahesvara do not simplify the case (SM 111:89,1-2): dasyateh 'tasu upaksaye dasu cd* ity asya ksaydrthe vartamdnasyeti sesah,' "dasyateh", [that is to say,] of this [about which it is stated Dhp 4.106-7:] "[the root] tas [occurs] when wasting [is to be denoted], so also [the root] das"; it is of that which occurs when there is the sense of destruction, thus the statement is to be completed'. Here too we fail to fin<f any reading ksaydrthdt\ instead we meet with two attributes in the genitive case. On this evidence, I think there is reason to doubt the authenticity of the wording of the Nirukta here. (9.14) sankdh91 sacatehl sampurvdd vd kiratehl sankdh 'conflicts' is from sacatih (^sac 'accompany; suffer'); or if is from kiratih l 'scatter') preceded by [the preverb] sam. Again, we face an apparently clear-cut case of an attribute in the ablative case, although it is not met with in Durga's commentary. In the commentary of Skanda-Mahesvara we meet with three different analyses, but any trace of sampurvdd is missing (SM 111:156,1): sankdh sacater gatikarmanah sangacchante vd tatra yoddhdrah samklrtyante vd, 'sankdh is of/from sacatih which denotes the activity of moving; or else [they are so called because] warriors
94 95 96

The context is RV 1.59 6. That is, /:arafoz-interpretations as adhikarana and hetu, respectively So read with Rajavade for sasyani. 91 The context is RV 6.75.5.

The universe ofYaska

131

come together (sam^lgam) in it, or they [= the warriors] are praised (sarplkirt) [in it]'. This does not reflect the extant wording of the Nirukta very well either, and the analysis should therefore be dismissed as a later interpolation. (11.24) dga9S dnpurvdd gamehl dgah 'sin' is from gamih (^Igam 'go; move') preceded by [the preverb] dN (a). This appears to be a straightforward case of an attribute in the ablative case. Durga quotes the text as above and adds (DII: 1067,11): avasyam evaitad dgacchati kartdram ity dgah, 'most certainly it comes (d^garri) to the one who commits [it], thus [it is called] dgah\ The commentary of Skanda-Mahesvara is missing at this point. On this basis there is no reason to doubt the text here. The only point of uncertainty is that Yaska elsewhere (Nir 1.3) refers to the preverb d not by its technical name dN known from vydkarana, but simply by stating d iti. This material leaves us with eleven possible attributes in the ablative case to technical analyses of the -teh/-eh type. I would consider four of these possible but inconclusive, namely, vakter vd vahater vd sdbhydsdt (3.13), ardham harater vipantdt (3.20), hanteh . . . vipantdt (4.25), and latater vd sydd [lambakarmanah] viparltdt. Further, I would dismiss sarter abhyastdt (6.3) and sampurvdd vd kirateh (9.14), maybe also dasyateh ksaydrthdt (7.23), as later additions. This leaves us with four cases of attributes with an ablative case ending, three of which should be treated with some caution, and the four possible instances mentioned already. In other words, the material is not overwhelming, but, as in the case of the possible attributes with a genitive case ending, the important thing is that ablative endings do occur. It is evident from the above investigation that it is not possible to rule out either the genitive or the ablative interpretation of forms in -teh and -eh on the basis of possible attributes in the Nirukta itself. Further reason: etasmdd eva It is now necessary to investigate a formulaic expression which is frequently resorted to in the Nirukta in cases where a nirvacana analysis is followed by the analysis of a homonym or a similar word: x apy etasmdd eva. It would certainly not be fanciful to assume that if this expression occurs immediately after an analysis of the -teh type, it could refer back to it and thus give support to the interpretation that the forms in -teh are ablatives: 'x too is derived from this t [root]'. I do not, however, think that this is how the formula was intended in the i first place. On the contrary, it is perfectly possible that the formula is to be seen , in relation to the most fundamental aspect of nirvacana analysis: why is something called what it is called? In other words, it would link up with the interrogative kasmdt 'why?' 99 so often rhetorically used to introduce an analysis,
98

The context is RV 1.94.15.

" See Kahrs 1983.

132

Indian semantic analysis

and with formal analyses of the -andt type giving the reason why something is called what it is called. Accordingly, the formula 'x apy etasmdd eva' would have to be interpreted as 'x too [is so called] for the very same reason'. Since the overall tendency among scholars, notably Sarup (1921), Skold (1926) and Rajavade (1940), has been to take etasmdt as a pronoun referring back to a form in -teh which therefore would have to be a form in the ablative case, I deem it necessary to present in extenso the material from the Nirukta exhibiting this formula, and then see what conclusions can be drawn from a detailed investigation of it. (1.20a) bhimo bibhyaty asmdtl bhismo 'py etasmdd eval bhirnah 'fierce; terrible' [is so called because] one is afraid (Vfr/iF 'fear') of it; bhismah 'terrible; dreadful' too [is so called] for the very same reason. Can there be any doubt that this gives us a causal explanation? I believe not. Durga (D 1:151,9) merely quotes the text, not even finding it worthwhile to comment. The same is the case with Skanda-Mahesvara (SM 1:119,11) and Nllakantha (NSV 1.6.233). (1.20b) girih parvatah I samudgirno bhavatil. . . I meghasthdylI megho 'pi girir etasmdd eva I ginh means a mountain; [it is so called because] it is elevated (sam-uch\gf)\ . . . 'seated on a cloud': meghah 'cloud' too is [called] ginh for the very same reason [i.e., because that too is elevated]. Skanda-Mahesvara (SM 1:120,6-7) comment: megho 'pi girir etasmad eva, samudgxrnatvdt so 'pi hi samudglrna ivdntariksena, '"meghah 'cloud' too is [called] girih for the very same reason", [that is to say,] because of the fact that it is elevated, for that too is elevated, as it were, by the atmosphere'. There is no doubt that they take etasmdd eva to be a causal statement. Nllakantha is of the same opinion (NSV 1.6.151): udgirna via megho 'pi vdyund vd svarena vdI etasmdd itipurvoktam udgirnatvam uddhriam, 'a cloud too is as if elevated, by wind or by sound [?], and is called [girih] "for the [same] reason", that is, the aforementioned property of being elevated'. This refers back to samudgirnatvahetutah 'for the reason of being elevated' (NSV 1.6.238). (2.5) gaur iti prthivyd ndmadheyam I yad duram gatd bhavatil yac easy dm bhutdni gacchantil gdter vaukdro ndmakaranah I athdpi pasundmeha bhavaty etasmdd eval gauh is a name for the Earth, since it stretches out (^gam\ lit. 'is gone') far, and since living beings move (^gam) on it; or it is of/from gdtih (^Iga 'go' or 'sing; praise' 10°), ioo £>urga (D i 179J) takes it to be Vga 'move', that this root occurs only in the dtmanepada as a second class root is no indication that he is wrong since Yaska often treats roots occurring only in the dtmanepada as a noun in -ft, it seems to have troubled Skanda-Mahesvara, though, who stick to ^Igd 'sing, praise' (SM 11*42,12)

The universe of Ydska

133

[and] the syllable o is a noun-maker;101 moreover, it is a name for a tethered animal here, for the very same reason. Here it is possible that etasmdt refers back to gateh. It is, however, perfectly possible to stick to the interpretation I have suggested, and understand Yaska to say that a cow is called gauh for the very same reason that the Earth is called gauh, because it moves and because it is praised. But the commentators complicate the picture here. Durga states (D 1:179,8-9): etasmad eva dhdtudvaydd gamer gdter vd. The only natural way to interpret this is: 'etasmad eva, that is to say, from these two roots, namely ^Igam or ^gd\ Skanda-Mahesvara (SM 11:43,1) too add dhdtudvaydt after etasmad eva. It is of course possible to interpret dhdtudvaydt as 'because there are two roots', and Durga's statement to mean: ' "for the very same reason", that is to say, because there are two roots it is either of ^Igam or of Vga'. Admittedly this would be a bit forced. But something does not make sense here. Clearly, yat is in both instances to be understood as yasmdt 'since; because'; so also Durga (ibid.,5-6). This makes it natural to interpret etasmad eva as 'for the very same reason(s)' which would link it to the two alternatives first presented. Skanda-Mahesvara (ibid.) take iha 'here' in the sense of iha loke 'here in this world' and add, to be on the safe side, vede ca 'and also in Veda'. Durga, on the other hand, (ibid.,8) explains iha as kdrakadvaye, 'where a two-fold [interpretation with regard to] kdraka [is possible]'. This refers directly back to the alternative interpretations as the agent of the act of going ('it stretches far') or the locus of the act of going ('living beings move on it') which he has already presented us with (ibid.,6—7): evam kartrkdrakam adhikaranam vd yojyam, 'thus it suits [to interpret as] the agent-participant or the locus'. It is possible, then, that the commentators simply quote the familiar expression etasmad eva but consider it necessary to add dhdtudvaydt 'from these two roots', namely Vgam and Vga, since there is some confusion here as to what Yaska had in mind in as much as an interpretation as adhikaranasddhana is a bit difficult in the case of a cow. There is, however, nothing to determine what Yaska had in mind, and the passage as it stands does not very easily lend itself to any one particular interpretation. (2.6) vir iti sakunindmal veter gatikarmanah I athdplsundmeha bhavaty etasmad eva vih is a name for bird: it is of/from vetih (Vvf) which denotes the activity of moving, moreover, it is a name for arrow here, for the very same reason. Once again it is possible that etasmad eva refers back to veteh which would |hus have to be an ablative form, but it is by no means necessarily the case. As |he text stands, the context is vayah 'birds' from RV 10.27.22, but there is *fvery reason to be suspicious of the text here since the passage is not quoted <br commented upon by Durga or Skanda-Mahesvara in the customary way.
M

See pp. 115 f above for a discussion of gdter vaukaro ndmakaranah

134

Indian semantic analysis

I am therefore inclined to dismiss it as a later addition, and from that angle it is possible to argue that in both this and the preceding passage we>are dealing with later interpolations by someone who sloppily thought etasmdd eva was used to refer back to forms in -teh, that is to say, by someone who was only interested in the formal grammatical derivation. (2.8) yonir antariksaml mahdn avayavah I parivito vdyundl ayam apitaro yonir etasmdd eva I pariyuto bhavatil yonih means atmosphere: it is a vast part (avayavah: ava^lyu); [it is] encompassed (parivitah),102 [that is to say,] by wind (vayund)', this other yonih 'womb' [is so called] for the same reason: it is embraced/surrounded (parblyu). This does not add anything specific; it is of course possible that etasmdt refers to the element yu that is part of the formation of avayavah and then repeated by the element yu that is part of pariyutah. This is how Samp (1921:27) interprets: This other (meaning, i.e.) a woman's womb, is derived from the same root also: it is surrounded.' I do not find this convincing, though, since the nirvacanas are expressed;n a non-technical language. Instead, I think it refers back to the meaning of being encompassed, embraced, surrounded. This seems to be how Durga too understands the situation (D 1:191,1-2): ayam apltarah strlyonir etasmad eva/ asdv api pariyuta eva bhavati sndvnd mdmsena ca, '"also this other \yonih\\ that is, a woman's womb, [is so called] "for the very same reason", for this too "is surrounded", that is, by sinews and flesh'. Skanda-Mahesvara remark as follows (SM 11:63,5): itarah strlyonir etasmad eva rupasdmdnydt prasakto nirucyate,' "this other", namely a woman's womb, [is so called] "for the very same reason"; it is incidentally analysed because it is identical in form'. This is how I understand this remark, which, then, tells us nothing. The other possibility is that rupasdmdnydt 'because it is identical in form' glosses etasmdt, but I find this less than likely. In other words, the commentators are not very explicit here, and the instance remains unclear. (2.13) vratam iti karmandmal nivrttikarma vdrayatiti satah/ idam apitarad vratam etasmdd eva vrnotiti satah/103 'vratam' is a name for an action; [m its sense of] act of abstention (i.e., vratam in the sense of 'vow') it is of that which really exists so that one says [of it]: it restrains (Vvr 'cover; hide', caus. 'restrain');104 this other vratam too (i.e., vratam in the general sense of 'action') [is so called] for the very same reason: it is* of that which really exists so that one says [of it]: it covers [you] (Vvr 'cover'). Here it is quite impossible that etasmdt refers back to vdrayatiti satah sincq satah here and elsewhere in constructions of the same type has been proven tq
102 103 104

The context is RV 1.164 32 = AV 9 10 10 where the word panvftah occurs. See Kahrs 1980.132, 234-6 A more exhaustive way of interpreting an analysis of the iti satah type will be argued for later,

The universe of Yaska

135

be a genitive form.105 Rather, it refers to the reason why vratam is so called: it causes to be enveloped. The only difference between the two analyses is that in the first we have a causative and in the second a present indicative form of the same root Vvr 'cover': vratam in the sense of nivrttikarma, that is to say, 4 vow; vow of sexual abstention', keeps you back, restrains you (vdrayati), while vratam in the sense of karma, action in general, envelops you and binds (vrnoti) you with regard to the consequences. The context here is RV 1.24.15; vratam is listed under karmandmdni Nigh 2.1. Although it does not seem to be of importance from the present point of view, it ought to be mentioned that Durga (D 1:207,15-208,2) reverses the analyses. That is to say, he gives the text as vratam iti karmandma vrnotlti satahl idam apitarad vratam etasmdd eva nivrttikarma vdrayatiti satah. Skanda-Mahesvara too (SM 11:84,18-85,4) support such a reading, which would indeed be preferable on the grounds that one would then move from vratam in the general sense of 'action' indicated in the Nighantu to the more special sense of 'vow'. However, this situation does not affect the interpretation of the etasmdd eva construction. Here it has to mean 'for the very same reason'. Sarup (1921:30) translates the passage as follows: 'The word vrata is a synonym of action, having the sense of abstaining: because it enjoins. This other meaning of vrata (i.e. a vow) is derived from the same root also: because it chooses.' This may serve as a good illustration of how blind Sarup is to the nuances of the various types of technical analyses. So is Skold, who remarks (1926:331): 'The words vrmtiti [sic] sato contradict the preceding statement, or else etasmdd eva are out of place.' Their confusion is not diminished by their taking vrnoti as a form of Vvr 'choose'. Rajavade (1940: Notes, p.357) does not confuse the roots, but remarks: This {idam) other (itarat) vrata also comes from this (etasmdt) very (eva) root; a performer of the Soma-sacnfice, for instance, binds himself to perform certain acts; these acts bind him from first to last; he must carry them out according to prescription; vrata in this sense comes from vr (=to cover, to bind). Apart from his taking etasmdd eva to mean 'comes from this very root', Rajavade gives a long list of reasons why vratam is so called. (2.17) ahir106 ayandtl ety antarikse I ayam apitaro 'hir etasmdd eva I nirhrasitopasargah I dhantlti I ahih 'cloud' [is so called] on account of [its] moving (ayanam: VJ): it moves in the atmosphere; this other ahih ('serpent') too [is so called] for the very same reason; [or?] with the preverb shortened: 'it kills' (d^lhan). This seems to be straightforward: ahih in the sense of 'cloud' is so called because it moves (eti), and ahih in the sense of 'serpent' is so called for the *05 See Kahrs 1980 130-48, 233-50, and 1984 150-2
106

Nigh 1.10.

136

Indian semantic analysis

very same reason, that is, because it moves. This interpretation is supported by Durga who simply adds (D 1:219,11) asdv api hy eti, 'for it too moves'. SkandaMahesvara, on the other hand, complicate matters when they state (SM 11:95,11-12): ayam apltaro 'hih sarpa etasmad eva eteh I asdv api hi bhumydm eti. Here etasmad eva is followed by eteh which could be a gloss. In that case Samp's 'from the same root also' would be correct. It is, of course, also possible to think of eteh as an added formal analysis without its having to be a gloss, in which case it does not give us any information on how etasmad eva is to be understood. That this is the case will become clear below. The third analysis met with in tjie Nirukta, as a derivation from orphan with the preverb shortened, comes rather abruptly and without indicating it as an alternative as is commonly done. Both Durga (D 1:218,11) and SkandaMahesvara (SM 11:95,12) add atha vd and yad vd 'or else', respectively. Durga (ibid.,12) rephrases this alternative as: dnpurvasya hanteh, 'of tiantih (^Ihari) preceded by dN (d)\ This clear display of a genitive case ending excludes the possibility that Durga would consider etasmad eva equivalent to eteh. (2.22) parjanyo vdyur ddityah sitosnavarsair osadhih pdcayantil anupd101 anuvapanti lokdnt svena svena karmandl ayam apltaro 'nupa etasmad eva I anupyata udakenal Rain, Wind, and Sun: with cold, heat, and rain they ripen plants; [so these three] anupah 'sowers'108 [are so called because they] sow (anuvapanti: artwlvap 'scatter; sow') the worlds with their respective activities; this other anupah ('watery place; marshland; shore or riverbank'109) too [is so called] for the very same reason: it is sown/scattered (anupyate: amAvap) with water. This would seem straightforward were it not for the fact that Durga remarks on etasmad eva in the following manner (D 1:229,1-2): vapaterdhdtor asdv api hy anupyate 'nuprakiryate110 nityakdlam udakena, '[that is, It is] of/from the root vapatih for this too "is scattered", that is, perpetually thrown at "with water"'. This, of course, makes it possible to argue that Durga actually glosses etasmdt by vapateh dhdtoh, but this would not be in keeping with the fact that Durga invariably treats of the forms in -teh as genitives. It is equally possible that Durga holds that etasmad eva means 'for the very same reason' and then adds the formal analysis. Skanda-Mahesvara do not quote the words etasmad eva but add (SM 11:106,7) vaper eva which clearly is a reference to ^Ivap and the added eva may indicate that it is linked to reflect the expression etasmad
107 108

110

The context is RV 10.27.23. This is how Yaska understands anupah in RV 10.27.23. The word occurs only twice in the Rgveda. In the present context Geldner translates anupah by 'Buffer vpr 'am Wasser wohnend', in the context of 9.107.9 by 'Marschland'. Yaska's derivations of anupa have been dealt with in detail by M.A. Mehendale (1965/78). For other views on this word, see the references in VB III; R. Turner (1979), in a review of Mehendale 1978, supports the derivation of anupd- from dp- 'water' by adducing examples of related words from middle and modern lan109 guages (Pali, Prakrit, Hindi, Sinhalese). So Durga (D 1:229,1). vX.prakiryate.

The universe of Yaska

137

eva. Once more, this makes it possible, but by no means certain that vapeh is intended as a gloss of etasmdt. In other words, on the basis of the Nirukta passage itself, it seems perfectly reasonable to interpret etasmdd eva as 'for the very same reason', but the remarks of the commentators no longer make this absolutely straightforward. The next passage to be discussed, however, offers a good reason for ruling out the possibility that vapeh in the commentary of Skanda-Mahesvara is a gloss of etasmdt. (3.9) dhurm dhurvater vadhakarmanah I iyam apitard dhur etasmdd eva I vihantivahaml dhuh 'finger; pole' is of/from dhurvatih (^dhurv) which denotes the activity of hurting;112 this other dhuh ('yoke') too [is so called] for the very same reason: it hurts the shoulder [of the ox]. Skanda-Mahesvara (SM 11:142,15) replace vadhakarmanah by vadhdrthasya, genitive. This effectively rules out the possibility that etasmdt should pick up dhurvateh. Durga, on the other hand, adds dhdtoh vadhakarmanah after etasmdd eva (D 1:274,11-12). This leaves us with the possibility that he intends to gloss etasmdt by these words, although I find it more likely that he adds these words as an explicit formal analysis, a conclusion supported by Durga's remarks on the following passage. (3.10) khala iti samgrdmandma1131 khalater vdl skhalater vdl ay am apitarah khala etasmdd eva I samdskanno bhavatil khalah is a name for battle: either it is of/from khalatih (^Jkhal 'tremble; shake'), or it is of/from skhalatih (^skhal 'stagger; stumble; fall'); this other khalah ('threshingfloor; granary') too [is so called] for the very same reason: it is scattered over. Since alternative explanations are given here, it seems feasible that if etasmdt were a pronoun referring back to a form in -teh, it can in this case only refer back to the last one, namely skhalateh; on the assumption that etasmdt gives a reason, this would also seem to be the case. Durga comments on the passage as follows (D 1:281,14-17):
v

khalater va bhrasyatyarthasya bhrasyanti hi tatra yodhdhl skhalater va himsdrthasya himsyante hi tatra parasparena I ayam apitarah khalah dhdnyakhalah etasmad eva tatrdpi hi bhrasyanti cumyamdndni dhdnydni himsyante va curnyanta ity arthah I atha va samaskannah asau bhavati viprakirnaih114 dhdnyaih I Yaska is commenting on Nigh 2 5 anguhndmani 'names for finger' where dhurah is listed. As his nigama Yaska quotes RV 10.94.7.

111

n }113 Nigh 2.19 lists dhurvati under vadhakarmanah. " Nigh 2.17; as his nigama Yaska quotes RV 1O'.48.7. m

So read for viprakirneh; ed. Rajavade (1921) reads viprakimo with v.l. viprakirnaih, Bhadkamkar (D 1.281, note to text) states 'a. d. viprakimo for viprakirnaih ofbi.' which indicates that viprakirneh is a mere misprint.

138

Indian semantic analysis

'Either it is of khalatiK which has the sense of falling,115 for warriors fall in it; 'or else it is of skhalatiK which has the'sense of injuring,116 for warriors ate injured by each other in it. This other khalalv too', [that is to say,] a threshing-floor, [is so called] 'for the very same reason', for on that too grain which is being ground falls; or else it is being injured, that is, ground - that is the meaning. Or else [it is so called because] it 'is scattered over', [that is to say,] with grain that has been thrown about. This passage is quite illuminating. First of all, Durga takes khalateh and skhalateh to be genitive forms. This rules out that etasmat refers back to any of them. Secondly, it is not only clear that etasmdd eva gives the reason, but that Durga lets it refer back to both alternatives: battle is called khalah because warriors fall in it, threshing-floor is called khalah because grain falls on it; or battle is called khalah because warriors are injured in it, threshing-floor is called khalah because grain is injured or ground on it. Finally, Durga takes samdskanno bhavati as a separate analysis, which I find surprising. The same position is taken by Skanda-Mahesvara, as will be clear below. The explanation could be that the meanings of Akhal and Askhal seem to present some difficulty to the commentators who cannot make any of them fit the meaning of samdskannah. Skanda-Mahesvara comment on this passage in an equally illuminating manner (SM 11:149,7-9): khala iti samgramanama, khalater va mathandrthasya, mathyante hi tatra yodhah, skhalater va, skhalanti hi tatra yodhah I ayam apityddi, etasmad eva khalateh/ tatrdpi hi sammathyante ksudyante sasydnil skhalantiti va, skhalanti tvaramdnds tatra karsakdh I samaskanna iti nirvacandntaram I 'khalah is a name for battle: either it is of khalatiK which has the sense of hurting, for warriors get hurt in it; 'or it is of skhalatiK, for warriors stumble in it; 'also this' etc., 'for the very same reason' is of khalatih, for on this too [something] is crushed to pieces, that is, trampled upon, namely corn; or else '[they] stumble', that is, ploughmen who move with speed stumble on it; samdskannah ('it is scattered over') is a separate analysis. It is quite obvious that Skanda-Mahesvara consider khalateh to be a genitive form and that they add it after etasmdd eva. If we demand just a minimum of consistency within one and the same paragraph, the only reasonable conclusion one can draw from this is that when they add a form in -teh after etasmat, here and elsewhere, they do not intend the form in -teh as a gloss of etasmat but as a repetition of the formal analysis - in the genitive case, u (4.5) musow musikd ity arthahl musikdh punar musndtehl muso 'py etasmdd eva I
115

116 117

Dhp 1.578, however, gives khala samcaye (calane); bhrasyatyarthah is not known from any dhdtupdtha. Skanda-Mahesvara (SM II: 149,7) give the meaning mathanam 'whirling around; shake; churn; hurt; destroy' to ykhal, this is not met with in any extant dhdtupdtha either. Dhp 1.577, however, gives skhala samcalane (samcaye ca)\ no dhdtupdtha gives skhala himsdydm. Nigh 4.1.1 take the form to be the nom. pi. of mils-. Sayana quotes the analysis at RV 1.105.8 where the form musah clearly is nom. pi.

The universe of Yaska

139

musah, that is to say, musikdh 'mice'; musikdh, again, is of/from musndtih (Vmws * steal'); musah too [is so called] for the very same reason. Here it is certainly imaginable that etasmdt refers directly to the analysis musndteh, however this is not how Durga takes it (D 1:364,2-4): adhundkhydnaprasaktasya musikdsabdasya nirvacanam karoti - musikah punar musnateh/ 'musa steye" ity asyal musah api ca etasmad eva te hi haranti dhdnyddinil Now he makes an analysis of the word musikd 'mouse' which has come up [because it has been used] in the explanation [of musah\: 'musikdh, again, is of musndtih\ [that is to say,] it is of this [which has been listed Dhp 9.58 as] '[the root] mus [occurs] when [the sense of] stealing [is to be denoted]'; and 'musah too [is so called] for the very same reason', for they snatch such things as grain. This leaves no room for doubt: musndteh is taken as a genitive form, and etasmad eva is taken to indicate the reason why musah is called musah. Skanda-Mahesvara, on the other hand, comment as follows (SM 11:208,17): tathd cdyarn muso 'py etasmad eva dhdtos tathokiah. It is perfectly possible to translate this as 'and thus also this [word] musah is in that way stated as from this very [same] root'. Here it would seem that dhdtoh picked up etasmdt directly. It is not quite certain that this is the way to interpret it, though, since Skanda-Mahesvara at the previous passage (3.10) added a formal analysis in -teh, considered to be a genitive form, after etasmdt. One could then take dhdtoh in the present passage as a genitive form and translate: 'and thus also this "musah [is so called] for the very same reason", [that is,] it is of the [same] root as stated'. In any case, the passage demonstrates clearly the very intimate link between a formal analysis and the reason why something is called what it is called. (4.8) madhum somam ity aupamikam mddyatehl idam apitaran madhv etasmad eva I madhu ('mead'?), metaphorically meaning Soma, is of/from mddyatih {^\mad 'be * exhilarated; exhilarate'); this other madhu ('honey; water'?119) too [is so called] for the very same reason. It is not entirely clear what senses of madhu Yaska has in mind, but this is not of primary importance here.120 Durga (D 1:371,1) adds mddyateh after
119 The context is RV 10.71.2. madhu is listed under udakanamdni Nigh 1.12. The first word madhu is necessarily the one used metaphorically of Soma. Indeed, the two words often occur side by side in the Rgveda as in the verse in question here (3.47.1), and madhu therefore seems to refer to an intoxicating drink in this context; cf. H. Luders (1959:343 f.): 'Ebenso emdeutig ist die Bezeichnung des madhu als "geprefit" (sutd), der Prefisteine sowohl als somasut wie als madhusut, und endhch das standige Nebenemanderg stehen von soma und mddhu, oft in demselben Verse. Aber gerade ein Ausdruck wie somydm u ** madhu scheint mir zu beweisen, daB madhu ursprunghch emen andern Trank bezeichnete und der Name nur auf den Soma ubertragen ist.' By the 'other' madhu I think Yaska intends r simply honey, which m itself is considered to possess intoxicating qualities, and not neces3 sarily any intoxicating drink. This is not quite how the commentators take it though. Durga 118 120

140

Indian semantic analysis

etasmdd eva, Skanda-Mahesvara (SM 11:218,2) add dhatoh. Again it would seem that these were added as glosses of etasmdt. One should also see the present Nirukta passage in the light of another analysis of madhu offered by Yaska at Nirukta 10.31: madhu dhamater vipantasya, 'madhu "honey"121 is of dhamatih {^dham 'blow; melt') which has been inverted'.122 This is one of the few apparently genuine instances in the Nirukta where Yaska identifies a form such as dhamateh as a genitive form. (4.13) api vd sira ddityo bhavatil yad anusete sarvdni bhutdnil madhye caisdm tisthatil idam apitarac chira etasmdd eva I samdsritdny etad indriydni bhavantil Or else sirah 'the head'123 is the sun, since it goes to rest along (anu^lsi) with living beings, and it stands in their middle; this other sirah ('head of a living being') too [is so called] for the very same reason: the sense-faculties [rest=] are dependent upon (sam-d^sri) it. I find this passage puzzling, and not very illuminating for the present investigation. The commentators are not very helpful here either, and one may suspect that they did not quite see what the underlying idea may have been. Skanda-Mahesvara (SM 11:226,19) point out that in the case of the second word sirah we are dealing with the human head, and again they add dhatoh after etasmdd eva. They continue: tad api anupravisya caksurddibhutdnindriydni tisthanti, 'having entered into it, the sense-faculties consisting of the eye etc., remain [there]'. This would indicate that the reason referred to by etasmdt is in fact madhye tisthati 'it stands in [their] middle'. It seems in any case clear that samdsntdni etc. is a statement intended to give a reason, and not a separate analysis. This interpretation is supported by Skanda-Mahesvara (§M 11:227,2) who refer to it as an upapattivacanam, 'a statement of evidence'. What is not clear to me is the following. If etasmdd eva were to indicate that the human head is called sirah for the very same reason that the sun is called sirah and the reason indicated for that is anusete sarvdni bhutdni, clearified as madhye tisthati, then we are facing some problems. First of all, whether anusete be transitive or not, it has to identify the sun as the agent of the activity expressed by that verb, whereas the human head is the karman, the object of the activity carried out by the senses. If the same reason is hard to find, one would have to consider the possibility that etasmdtreally refers only to a formal analysis here, that is to say, that it refers back to anu^lsT, the meaning of which is not very obvious in the context and the formal reference to which is much too vague for it to be picked up simply by the etasmdd eva construction. Footnote 120 (cont) (D 1-371,1) identifies it as maksikam ('honey' according to MW), and Skanda-Mahesvara (SM II 218,1-2) as maksikam, bhrdmaram ('bee-related, honey' MW), or ksaudram ('honey, species of honey L , water L' MW) 121 The context here is RV 4.38 10, a Dadhikra hymn, where it is stated that the words should be mixed with honey,m Renou (EVP XV* 164) 'impregne de miel ces paroles-ci\ and Luders cf. 123 (1959.347-8). Cf p. 122 above. The context is RV 1.163.10.

The universe of Yaska

141

To take etasmdd eva as referring to the derivation from a verbal root does not make much sense here. By way of example, Sarup (1921:60-1) translates: 'Or else siras refers to the sun, i.e. it follows all created things to rest, and stands in their midst. This other (meaning of) siras (i.e. the human head) is derived from the same root also: the senses depend upon it.' S. Varma (1953:92) is more explicit in indicating which verbal root he has in mind: ' "the sun" or "head", is traced to w " t o lie", lit. the former, because "it lies among all creatures", the latter, because "the organs of sense rest i.e. depend upon it"'. (4.16) upddarsi sundhyuvahl sundhyur ddityo bhavatil sodhandtl tasyaiva vakso bhdsddhyuhlam I idam apitarad vaksa etasmdd eva I adhyuhlam kdye I sakunir api sundhyur ucyate I sodhandd eva I udakacaro bhavati I dpo 'pi sundhyuva ucyante I sodhandd eva I '[Her breast (vaksah)] lets itself be seen [like that] of the sundhyU'124 (RV 1.124.4); sundhyuh is the sun, [so called] on account of the cleansing (sodhanam: ^sudh, caus.); its very breast (vaksah) is raised/abundant (adhyudhah: adhi + ^lvah 'carry') with light; this other vaksah ('breast [of human or animal]') too [is so called] for the very same reason: it is raised on the body; a bird too is called sundhyuh on account of the very same cleansing, [for] it lives in water; the waters too are called sundhyuvah125 on account of the very same cleansing. Note that the explanations with a form in -andt, clearly causal explanations, are exactly parallel to the etasmdd eva construction; when they are repeated they even contain the word eva. This supports the hypothesis that etasmdd eva has to be interpreted as 'for the very same reason'. To assume that etasmdt were to refer to a well concealed verbal root here would be far-fetched. Sarup, missing out one sentence, nevertheless translates (1921:62): 'The pure one is the sun, (so called) from purifying. This other (meaning of) vaksas (breast) is derived from the same (root) also: it is exalted in the body.' Durga (D 1:388,5) adds vaher dhdtoh and Skanda-Mahesvara (SM 11:237,2) add dhdtoh after etasmdd eva\ in both commentaries this is followed by an expanded version of the reason given by Yaska. Once again, I think they feel a need to explicate because the analysis in the Nirukta is a bit cryptic and so they add a formal analysis or at least point out the connection with the preceding one.
124

The context is RV 1124 4 upa adarsi sundhyuvo nd vdkso° The hymn is to Usas Here the form is gen sg., in the feminine. The meaning of the word is not quite clear Geldner translates* 'Es zeigt sich lhre Brust wie die erne Sundhyuvogels', and remarks in a note* 'sundhyu nach Say. em weiBfarbiger Wasservogel, nach Durga der Madgu Auf ein Tier weist auch 8, 24, 24' Renou, on the other hand, translates (EVP 111*61). 'Sa poitnne se laisse voir comme celle de la courtisane', and remarks in a note (ibid..63) 'Sundhyu nom d'animal possible, surtout s'll est perrms de considerer aussi comme n d'animaux admasdd et nodhds Noter qu'au Livre X (138, 5), par possible meconnaissance de notre passage, U. elle-meme est appelee sundhyu Le mot est aisement etymologisable comme ep quahficative et se dit au fern d'une jeune femme X 39, 7, des juments du char solaire I 50, 9, de la pensee poetique VII. 88, 1 La traduction par "courtisane", mcertaine en soi, vaut en fonction de rensemble des hy. a U' 125 The form must be nom pi. in the feminine. Rajavade (1921; 1940) reads sundhyavah

142

Indian semantic analysis

(4.18) kacchapo 'py akiipdra126 ucyate I akupdro na kupam rcchatltil kacchapah kaccham pdtil kacchena pdtiti vdl kacchena pibatiti vdl kacchah khacchah khacchadahl ayam apitaro nadikaccha etasmdd eval kam udakaml tena chddyatel A tortoise too is called akupdrah, [that is to say,] a-kupa-arah: it does not (a = na) move (ara-\rcchati) in a well (kupa-)\ kacchapah 'tortoise' [is so called because] it protects (^pd 'protect') its mouth-and-neck (kacchah), or [because] it protects with its mouth-and-neck, or [because] it drinks (^pd 'drink') with its mouth-andneck;127 kacchah ('mouth-and-neck of a tortoise'), [that is to say,] kha-cchah, that is, kha-cchadah 'which covers (chddayati) space (khani)'; this other river-kacchah ('river-bank') [is so called] for the very same reason: kam, that is, water - it is covered (chddyate: ^Ichad 'cover', pass, caus.) by it. Again, it is perfectly reasonable to interpret etasmdt causally, in line with the other explanations given in this passage. Durga supports this interpretation when he adds after etasmdd eva (D 1:402,20): kasamjnakena udalcena chddyate Hi kacchah, 'by means of the thing called ka9 that is, by water, it is being covered, thus [it is called] kacchah9. (4.19) harihm somo haritavarnahI ayam apitaro harir etasmdd eval harih means Soma, [so called because] its colour is tawny; this other harih ('monkey') too [is so called] for the very same reason. Clearly, etasmdt gives a reason. Durga supports this (D 1:417,1): ayam apitaro harih markatah etasmad eva iti haritavarnatvdd dharir ucyate, ' "also this other harih", that is, a monkey, "for the very same reason", that is, it is called harih because of the fact that its colour is tawny'. So do SkandaMahesvara (SM 11:255,19): etasmad eva varnasdmdnydt, '"for the very same reason", that is to say, because the colour is the same'. ; (4.21) bandhuh sambandhandtI ndbhih samnahandtl ndbhyd sdnnaddhd garbhdjdyantel ity dhuhl etasmdd evajfidtTn sandbhaya ity dcaksatel sarnbandhava iti ca/jfidtih samjndndtl bandhuh 'tie; nexus; [maternal] relative' [is so called] on account of the binding together (sambandhanam: sant\bandh)\ ndbhih 'navel; umbilical cord; binding' [is so called] on account of the tying together (samnahanam: sam^nah); it is said: 'foetuses are born tied by the umbilical cord'.129 For this very reason one refers to close relatives (jndtayah) as sa-ndbhayah ('who have the same ndbhiK) and also as sambandhavah ('who have the same bandhuh'); jhdtih 'close relative' [is so called] on account of [his/her] being well known (samjhdnam: sam^ljfid).
126 127

129

The context is the word dkuparasya RV 5.39.2. Most scholars take kacchah to mean 'mouth' or 'shelF as it suits them. Skanda-Mahesvara (SM 11:247,6) claim the kacchah of a tortoise is located on its back (kacchapasya prsthe yat kaccha ucyate0). m The context is RV 10.101.10. The quotation has not been traced; Sarup'( 1927:75, note 6; 1921:66, note 9) remarks: 'Cf. TS. vi. 1. 7. 2.' The resemblance is rather slight.

The universe ofYaska

143

The context here is RV 1.164.33: dydur me pitajanita nabhir dtra bdndhur me mdta prthivimahiydm0'. Only pita, which is also analysed in this passage of the Nirukta, occurs at Nighantu 4.1 which is the passage presently being commented upon. Durga comments only on Yaska's analysis of pita apart from offering his own comments on the verse. This indicates that the passage may be a later addition. It is commented upon in full by Skanda-Mahesvara who explain etasmdd eva by sannahandt sambandhandc ca, 'because of the tying together and the binding together' (SM 11:260,16). Even though there is reason to doubt the text here, the passage nevertheless provides a clear example of how etasmdd eva can be used at the beginning of a clause as a clear statement of cause. This is also how Skanda-Mahesvara take it. (4.27) eko 'svo vahati saptandmddityah1301 saptdsmai rasmayo rasdn abhisanndmayanti I saptainam rsayah stuvantiti vdl idam apitaran ndmaitasmdd eva I abhisanndmdt I One horse with seven names (sapta-ndman-) draws: that is, the sun [draws]; seven rays bring the juices to bow towards (abhi-sanAnani) him, or seven sages praise (ynam in the sense of ^Istu) him; this other ndma ('name') [is so called] for the very same reason, that is, on account of the bending towards {abhisanndmdt: abhisarrnnam). Here Yaska himself adds a word expressing a cause after etasmdd eva. Durga remarks (D 1:445,9-10): idam apltarat sabdah nama etasmad eva abhisamnamat tad api svam arthatn pratydyayitum kriydpadasya vd gunabhdvendbhimukhyena samnamatil 'Also this other' word 'ndma [is so called] for the very same reason, that is, on account of the bending towards', [that is to say,] that too bends towards (directs itself) in order to make its own meaning recognised or [that] of a word denoting an action, directly as a subordinate.131 Skanda-Mahesvara add (SM 11:281,18-19) natam hi svasmin ndmavati, 'for it is bent towards its own name-possessor'. This, then, is an instance where both the Nirukta and the commentators make it absolutely clear that etasmdd eva gives a causal explanation. (5.1) tamo 'py andha ucyateI ndsmin dhydnam bhavatil na darsanaml andham tama132 ity abhibhdsante I ayam apltaro yndha etasmdd eva I
130 131

The context is RV 1.164.2. Rajavade (1921: tippanT, p. 148): vdkye dkhydtam pradhdnam itarat sarvam gaunaml e krtvd ndmdny api svdrthadyotane knydrthadyotane vd dbhimukhyena visesato guni samnamantil gumbhavantity arthah, 'in a sentence the verb is primary, everything else secondary; in recognition of this, nouns too, becoming especially subordinate through being directed to the manifesting of their own meaning or the meaning of an action, bow towards; that is to say, they become subordinate'. 132 Sarup (1927) reads andhantama0; Bhadkamkar 1918 and Rajavade 1921 read andhamtama0; in Samp's translation, however, we find (1921:72): 'People also use the expression andham tamas, i.e. "blinding darkness".' This is in agreement with the edition of Roth (1852). It seems

144

Indian semantic analysis

Darkness too is called andhah (andhas-):133 there is no (na = a) focusing (dhydnam: ^Idhya) on it, no seeing; one talks about blind darkness; this other andhah (andha'blind') [is so called] for the very same reason. This passage is followed by pdsyqd aksanvan nd vi cetad andhah, 'he who has eyes sees, he who is blind cannot know' (RV 1.164.16) as a nigama. This is quoted by Durga (D 1:458,5), but ayam apitaro 'ndha etasmdd eva is not. The analysis may therefore be a later addition, maybe even on the basis of Durga's commentary since he says (ibid.,7) itaras tu yo 'ndhah0, 'but that which is this other [word] andhah . . .' followed by a long explanation which need not concern us here. Skanda-Mahesvara, on the other hand, quote the analysis and add (SM 11:286,14) darsandbhdvdt, 'because there is no seeing', revealing both that they take etasmdd eva to allude to a causal explanation and that they identify that which is to be expressed by the preceding na darsanam in the Nirukta. The analysis may not stem from Yaska, but it is clear how Skanda-Mahesvara understand an analysis of this kind. (5.2) anyena mad ahano134 gaccha ksipraml dhamsiva bhdsamdnety asabhyabhdsandd dhand iva bhavatil etasmdd dhanah sydtl 'With someone other than me, O unchaste one' go quickly; speaking [in this way] you hurt (d^han) [me] as it were; [that is,] she is an dhanah 'unchaste woman' as it were, because of [her] vulgar talking; because of this it would be dhanah 'O ufichaste one'. This does not contain the typical set phrase with its ordinary syntax, but the underlying idea is the same, I believe. So does Durga (D:466,8): etasmad eva kdrandd ahanah syat°. Not only does he add eva to arrive at the familiar set expression, he even adds kdrandt 'for this reason' which is giving very explicit support to the interpretation of etasmdd eva as 'for this very reason'. Sarup, on the other hand, translates (1921:73): 'Go at once, O wanton,, with some person different from me. Speaking in this manner, thou hurtest me as it were. Ahand (a wanton) is (so called from) her lascivious speech. Ahanah (i.e. the vocative) is derived from the same.' This simply does not add up very well. (5.4) vardho135 megho bhavatil vardhdrahl varam dhdram dhdrsih136/ iti ca brdhmanaml vidhyad vardhdm tiro ddrim dstd1311 ity api nigamo Footnote 132 (cont.) to me preferable to take it as two words, particularly because Panini teaches the addition of the taddhita samdsdnta aC to the nominal stem tamas- when it is compounded with andhaas the first member, the resulting form being andhatamasa- 'blind darkness' (A 5.4.79 avasamandhebhyas tamasah). 133 andhah Nigh 4.2; vardhdh has already been explained as a name for food (andha ity annandma). 134 Both Sarup (1927) and Bhadkamkar (1918 = D 1:459) read ahano, but the nigarha offered here is RV 10.10.8 which would require dhario. The context is dhanah Nigh 4.2. 15 ^gh 42; the word is listed under meghandmdni Nigh 1.10. 3 136 The quotation has not been traced; Sayana quotes it, obviously with reference to the present passage of the Nirukta, at RV 1.61.7; Sarup (1927) notes that several MSS, all of them belonging to the longer recension, read it after etasmdd eva; Durga does not quote it. 137 RV 1.61.7. *

The universe of Yaska

145

bhavatil ay am apltaro vardha etasmdd eval brhati muldnil varam varam mulam brhatiti vdl vardhah is a cloud: it brings about the best (vara + d^lhr); there is also a Brahmana passage: 'You have brought about the best food/livelihood.' 'He pierced the cloud by throwing the thunderbolt across'; this too is a quotation from Veda. This other vardhah ('boar') [is so called] for the very same reason: it tears up the roots, or else [because] it tears up root[s] to its heart's content (varam varam) (i.e., varam mulam brhati [vrhati] = vardhdrah). The only natural interpretation is to take etasmdd eva in a causal sense. Durga supports this (D 1:479,17-18): asdv api hi varam muldkhyam dhdram dharaty eva, 'for this too does indeed fetch (dharati) [for itself] the best (varam) food (dhdram) known as roots'. So do Skanda-Mahesvara who add after etasmdd eva (SM 11:303,11): vardhdratvdd vardharanatvdd vd, 'because it brings about the best, or because it fetches the best'. (5.7) arir amitra rcchateh I Tsvaro }py arir etasmdd eval arihm [in the sense of] someone who is not a friend is of/from rcchatih (^Ir/rcch 'move; attack; invade; hurt'); a lord too [is called] arih for the very same reason. Here it seems possible that etasmdt refers back to rcchateh, and Durga (D 1:500,1) adds dhdtoh which could certainly be a gloss of etasmdt. SkandaMahesvara, on the other hand, remark as follows (SM 11:322,12-13): yathd arir rcchater gatyarthasya satrun prati gacchaty asau tair vd aryate I evam Isvaro 'pi tena tena bhutdnugrahena sarvapravrttir gatih I Just as (yathd)139 'arih ("enemy") is of rcchatih' which has the sense of moving, [because] he moves against his enemies or [because] he is moved [against] by them, in the same way (evam) a ruler too: through this and that favour to living beings, there is activity for all, that is, movement. : The addition of gatyarthasya makes it quite clear that etasmdt does not refer |to rcchateh: it indicates a cause, not a formal analysis. But the passage is a good Example of how both elements are involved: just as140 arih 'enemy' is of hcchatih because of this and that, in the same way arih 'lord' is of rcchatih for |tjie very same reason. Note, however, that it is the causal link which is primary; at is to say, etasmdt does not refer to rcchateh [gatyarthasya] but states the ason why there is a grammatical link between arih and rcchatih.

t
$

jg| (5.14) puskaram141 antariksamIposati bhutdniI udakam puskaramI | p pujdkaramlpujayitavyaml idam apitaratpuskaram etasmdd eval | g puskaram vapuskaram vdl

P

Analysed in passing. The editor (Sarup) remarks: '"yatha" ity asya padasya sthdne ca "yatah" iti pdtho yuktah sydt\ that is to say, 'instead of the word yathd, the correct reading should be yatah "since; \%i because'". This would of course give the whole passage a different meaning. I am not conp-1 vinced by this because of the following evam. 140 But see the note on yathd above. p The context is puskare RV 7.33.11; puskaram is listed under antariksandmdni Nigh 1.3.

146

Indian semantic analysis

puskaram means the atmosphere: it sustains (^pus) living beings; water is [called] puskaram [because] it is [a means of] performing (-kara) worship (pujd) [or because] it is to be worshipped; this other puskaram ('lotus') [is so called] for the very same reason; or else puskaram [is really vapuskaram, so called because] it creates (-kara) a beautiful appearance (vapuh). Once more, we are obviously dealing with an expression which suggests that puskaram 'lotus' is so called for the very same reason that water is called puskaram: either because it is a means of performing worship, or because it is the object of worship. Durga is quite explicit here (D 1:527,8-10): idam apltarat puskaram padmam etasmad eva kdrandt puskaram ity ucyate tad api hi pujakaram pujayitavyam ca sobhanatvdt tasyal 'this other puskaram', that is, a lotus, is called puskaram 'because of that very same', that is to say, reason; for that too is a means of performing worship and is [itself] to be worshipped because of its beauty. u Not only does he spell things out by changing it into the unambiguous expression etasmad eva kdrandt, 'for that very reason'; he also states what the reason is, namely the same two reasons that were given in the previous analysis. Skanda-Mahesvara (SM 11:345,12) add dhdtoh, and continue by giving the causal explanation (ibid.,13) tendpipujyate pujyate vd tad evam sobhanatvdt, 'worship is performed by means of it too, or it is [itself] worshipped in this way because of its beauty', explanations as karana- and karmasddhana respectively. (5.22) krttih142 krntatehl yaso vd/ annam vd/ mahiva kfttih sarana ta indra I sumahat ta indra saranam antarikse krttir iveti I iyam Izpitard krttir etasmad eva I krttih is of/from krntatih (^krt 'cut') [and means] either fame or food: 'O Indra, your protection is great like fame/food' (RV 8.90.6), [that is,] your protection in the atmosphere, Indra, is very great, like fame/food; this other krttih ('hide; skin; patched-up cloth') [is so called] for the very same reason [or: is from the same (root)]. Here etasmdt could easily refer back to krntateh, but taking it as indicating a causal explanation works without any problem: because of the cutting. Durga remarks (D 1:554,10-11): sdpy etasmad eva krntateh sd hi vastrdvayavaih krttair grathitd bhavati. This could be translated 'this [krttih 'patched-up^ garment'] too is from this, that is, from krntatih, for that is put together by means of parts of garment which has been cut'; but it could equally well bei translated 'for the very same reason the [krttih] too is of krntatih, for that is put together by means of parts of garment which has been cut'. In other words, it may work both ways. The possibility that we are dealing with ^krt 'weave; spin' as indicated by Varma (1953:130) is ruled out by the fact that this would
142

Nigh 4.2; kfttih also occurs Nigh 3.4 under grhanamani 'names for house';

The universe of Yaska

147

require krnatteh and not krntateh. Skanda-Mahesvara, obviously having the sense of 'skin; hide' in mind, add a causal explanation after etasmdd eva (SM 11:371,5-6): bahudhd krttd hi sd sarirdt, 'for this is cut in various ways from the body'. (5.26) kosah143 kusndtehl vikusito bhavatil ayam apltaro kosa etasmdd eva I sancaya dcitamdtro mahdn bhavatil kosah 'pail; bucket'144 is of/from kusndtih (^kus 'tear; force out; extract'): it is torn out; this other kosah ('treasure') too [is so called] for the very same reason: it is a great collection the size of what has been accumulated. Parallel to the previous instance Durga (D 1:566,2) adds kusndteh after etasmdd eva, and again I do not think there is any reason to assume, contrary to Durga's ordinary practice, that this is intended as a gloss on etasmdt. This conclusion is supported by Skanda-Mahesvara (SM 11:380,4): ayam apltarah kosa etasmad eva dhdtor vrddhyarthasyeti sesah, ' "this other kosah too for the very same reason": it is of the [same] root which has the sense of growth, thus [the analysis] is to be completed'. This makes it quite clear how they think matters work: etasmdt gives the reason why something is called what it is called; this establishes a link with a verbal root; to complete the analysis they sometimes add dhdtoh and sometimes a form in -teh depending on how obvious the grammatical analysis is; these are genitive forms and therefore not glosses of etasmdt as vrddhyarthasya makes evident. (6.1) sucih145 socateh I jvalatikarmanah I ayam apltarah sucir etasmdd eva I nihsiktam asmdt pdpakam iti nairuktdhl sucih 'burning; glowing' is of/from socatih (Viwc) which denotes the activity of burning; this other sucih ('pure') [is so called] for the very same reason; sin is removed {nihsiktam: nilplsic) from it, according to the Nairuktas. It is not quite clear whether the last part of this is intended as a further explanation of etasmdd eva or whether it is intended as a separate analysis in its own right. Durga believes in the latter alternative (D 1:575,15-18): sucih socater jvalaty arthasya/ ayam apltarah sucir laukika etasmad ever/ vaiydkarand manyante nairuktdh punar nihpurvdt sihcater itarah sucir ity evam manyante nih niskrsya asmat papakam asucitvam anyasmin siktam bhavatity evam I 'sucih is of socatih' which [denotes the activity of =] has the sense of 'burning; this other sucih\ the ordinary one, '[is so called] for the very same reason', [is what] the u. grammarians think; the Nairuktas, on the other hand, think that this other [word] sucih is [a derivation] from [the root] sic ('pour out') preceded by [the preverb] nih, thus: nih, that is, extracting 'sin from it', that is, impurity, it is poured (siktam) into something else.
143 .144 The context is RV 10.101.7. The word kosah is listed under meghandmdni 'names for cloud' at Nigh 1.10. 145 The context is RV 2.1.1.

148

Indian semantic analysis

I am not convinced that Durga has got this right, but the passage does in any case contain several interesting features. First of all, Durga makes it clear that he takes socateh to be a genitive form. This excludes the possibility that etasmdt refers back to it, but not that he does not take it to mean 'from the very same root {^suc)\ It is, however, perfectly possible that it indicates a reason, that is, that the reason for the word suci 'pure' is the same as the one for suci 'burning' and for that reason to be derived from the same ^suc. This analysis he attributes to the grammarians. The following is then considered a separate analysis which Durga makes formally clear by nihpurvdt sincateh, indisputably ablative forms. But the very fact that this is a separate analysis excludes the possibility that it picks up etasmdt which has already been linked to the genitive form in -teh preceding it. (6.17a) srprah146 sarpandtl idam apitarat srpram etasmdd eval sarpir vdl tailam vdl srprah 'slippery; oily; gliding' [is so called] on account of the gliding (sarpanam: V-srp); this other srpram (nt., 'ghee; oil') too [is so called] for the very same reason; ghee or oil. This illustrates well how the etasmdd eva formula is parallel to an analysis of the -andt type. Durga spells it all out (D 1:650,15-16): srprah (71) iti anavagataml sarpa ity avagamahl sarpanad iti nirvacanaml idam apTtarat srpram etasmad eva, sarpanad etat sarpir va tailam va, tad api hi sarpatil srprah (Nigh 4.3.71) is not understood; sarpah ('which glides') would provide an [immediate] understanding; sarpandt is the analysis; 'this other srpram too [is so called] for the very same reason', [that is to say,] on account of the gliding - 'ghee or oil' - for this too glides (sarpati). ^ Once more it is clear that we are dealing with a statement providing a causal explanation. (6.17b) skandho141 vrksasya samdskannam14* bhavatil skandha etasmdd eva I dskannam kdye I ayamnpitarah i

skandhah (skandhas- nt.) 'branching crown' of a tree [is so called because it] is attached {samdskannam: sam-d\skand) [to it]; this other skandhah (skandha- m.; 'shoulder') [is so called] for the very same reason: it is attached to the body. Again etasmdd eva clearly offers a causal explanation. Durga does not quote^ this analysis, he only comments upon RV 1.32.5 which is quoted as a nigama for the word kulisah which occurs at Nighantu 4.3. There is therefore every^ reason to assume that the present passage is a later interpolation. Skanda-l
146 148 147 Nigh 4.3. The context is skdndhamsi RV 1.32.5. Samp (1927) reads samaskanno0. This cannot be right. As is clear from the nigama quoted' (see preceding note), we are dealing with skandhas- (nt.); Skanda-Mahesvara too read: samdskannam0 (SM 11:456,11).

The universe ofYaska

149

Mahesvara comment upon it, however, and gloss (SM 11:456,12) etasmdd eva by dskannatvdt, 'because it is attached [to the body]'. This leaves no room for doubt: they assume that they are dealing with a causal explanation. (7.29) mithunau149 kasmdtl minotih srayatikarmd I thu iti ndmakaranah I thakdro vdl nayatih parah I vanir vdl samdsritdv anyonyam nayatahl vanuto vdl manusyamithundv apy etasmdd eva I methantdv anyonyam vanuta iti vdl Why mithunau 'couple'? [Because it is of/from] minotih (Vm/ 'fix') which denotes the activity of depending; thu is the suffix, or else it is the syllable tha, [with] nayatih (Vwf 'lead') or vanih (^van 'win')150 [to account for] the last [part]: depending on each other, they lead [each other] or they win [each other]. A human couple too [is so called] for the very same reason; or else [they are so called because] they win each other when they are uniting (methantau: Amith 'unite'). The last statement here offers an alternative explanation which contains an alternative verbal root to account for the first part of the word. This does not mean that it is not perfectly sensible to take etasmdd eva as offering a causal explanation. Skanda-Mahesvara (SM 111:110,3) add dhdtupratyayasamuddydt after etasmdd eva. If this is intended as a gloss, we are dealing with a case where they take etasmdt to refer directly to the formal elements of the analysis and not to the reason why something is called what it is called: 'from the [same] combination of roots and suffixes'. This, however, would not be in agreement with their own remarks at Nirukta 3.9 and 5.26 discussed above. (9.8) islkesater151 gatikarmanahl iyam apitaresikaitasmdd eva I isikd 'reed; rush' is of/from Tsatih (Vw) which denotes the activity of moving; this other isikd ('arrow') [is so called] for the very same reason [that is, because it moves]. This is a case where it would seem reasonable to interpret etasmdt as referring directly back to Tsateh, but the passage is almost certainly a later addition since it is an analysis of a word that has been used to explain a word that has been used to explain the form mdujavatasya which occurs RV 10.34.1, since it is not commented upon by Durga, since Skanda-Mahesvara quote it in the form iyam152 apTsikd apareslkd tasmdd eva (SM III: 149,5), and since the second word is isikd, not isikd. I am therefore inclined to dismiss this passage as a later interpolation. (10.44) budhnam153 antariksaml baddhd asmin dhrtd dpa [iti vd]154 idam apltarad budhnam etasmdd eva I baddhd asmin dhrtd prdnd iti I
149 151 152 153 15 The context is RV 10.88.11. ° Durga talks about sampmsarana here (D 1:833,8). The analysis is given in passing while commenting on RV 10.34.1. The editor notes: yam apisikdpa (pa C.) lesikd B. C. The context is dhir budhnyah Nigh 5.4. As a nigama Yaska offers RV 7.34.16 which also contains the word budhne. 154 Omitted by several MSS.

150

Indian semantic analysis

budhnam means the atmosphere, [so called because] waters are held bound (baddhdh: ^bandh 'bind') in it; this other budhnam ('body') too [is so called] for the very same reason: the breath is held bound in it. This passage quite naturally lends itself to be interpreted as giving a causal rather than a formal explanation. This is confirmed by Durga who adds very little here (DII: 1026,10-11): idam apftarat sanram budhnam ity ucyate,' "also this other one", that is, the body, is called "budhnam" \ This is simply followed by the words of the Nirukta, a fact which indicates that Durga felt no further explanation was needed; the Nirukta passage itself gives a causal explanation. Skanda-Mahesvara (SM IV:49,8-9) repeat Durga. (11.39) gaun155 rocateh I jvalatikarmanah I ayam apftaro gauro varna etasmdd eval prasasyo bhavatil gaun is of/from rocateh (^Iruc) which denotes the activity of blazing; this other gaurah too, that is, the colour, [is so called] for the very same reason: it is praiseworthy. This is not immediately clear. Durga remarks on it in the following manner (D 11:1085,12-14): gaurih156 rocateh ]\dldXyarthasyal mddhyamikd vdkl sd punah dlptimatil ayam apltarah yo gauro varnah so 'pi etasmad eva rocateh I sa hi dfptirupatvdt krsnddin apeksya prasasyo bhavati / *gaurih is of rocatiK which has the sense 'of blazing' [and is] Speech [as a deity] of the middle region: this, again, is bright; 'this other one too', the one which is 'the colour gaurah\ that one too 'for the very same reason' is of rocatih: for it 'is praiseworthy' with regard to such [colours] as black by fact of its being the colour of brightness. Durga's understanding of how the same causal explanation works is as good a guess as anyone's. It is possible here that rocateh glosses etasmdd eva but it is not very likely since Durga has already glossed rocateh with a form in the genitive case. The way to take this passage, I believe, is that Durga understands etasmdd eva to indicate a causal explanation and only secondarily a formal analysis, and that when the commentators add dhdtoh, which SkandaMahesvara do in this case (SM IV:86,10), or a form in -teh after etasmdd eva, these forms have not been intended as glosses on etasmdt, but have been added in order to clarify the formal analysis. 15 Nigh 5 5 As a nigama is offered RV 1164 41. The word gaun is listed under vanndmdni 5 'names for speech' Nigh 1.11. Geldner (RV 1164 41) takes gaurfas 'Buffel(-fell)', but see Luders (1951*252) who sees in gaurf 'allerdings den Himmelsstrom, aber zwischen diesem und der Vac bestehen die nachsten Beziehungen', and Renou EVP XVI 92; V S. Agrawala (1962) argues that gaun 'she-buffalo' symbolises Varuna's waters as well as vdk 'speech'. 156 This is how the word occurs RV 1164 41, clearly as nom sg , whether Durga actually read it like this in the Nighantu is hard to tell.

The universe of Yaska

151

(11.47) ano157 vdyur anitehl api vopamdrthe sydtl anasa iva sakatdd iva I anah sakataml dnaddham asmims civaraml aniter vd sydtl jivanakarmanah / upajivanty enatl megho }py ana etasmdd eval anah means wind [and] is of/from anitih (N'an 'breathe'); or else it may be [used in RV 4.30.10] in the sense of a comparison: as if from the anah, that is, as if from the carriage; anah means a carriage [and is so called because] rags are bound onto it; or else it is of/from anitih which denotes the activity of living, people make a living of it; a cloud too [is called] anah for the very same reason This is not particularly informative, and I am inclined to dismiss it as a later interpolation, first and foremost because anah sakatam and the rest of the passage is not met with in the commentary of Durga, but also because of the repetitive nature of the various analyses (aniter vd sydt jivanakarmanah) and their general mindlessness. Durga does quote the first analysis, ano vdyur aniteh, and makes it clear that he takes aniteh as a genitive form by glossing it prdndrthasya 'which has the sense of breathing'. This investigation leaves us with thirty-three instances of the etasmdd eva formula. Out of these I would dismiss five as later interpolations.158 Eighteen clearly indicate a cause.159 Five out of these eighteen are part of constructions which involve causal analyses of the -andt type.160 An atypical instance is 7.29 where various verbal roots and suffixes are given in the nominative case to indicate alternative derivations in reply to the question kasmdt 'why?' which one could easily argue is picked up again by etasmdt. In the remaining ten cases161 it is possible that etasmdt refers directly to forms in -teh or -eh, but all of these cases can equally well be interpreted as causal explanations. This warrants only one conclusion: in the first instance the formula etasmdd eva indicates a cause, not a grammatical derivation. All the occurrences met with in the Nirukta lend themselves to such an interpretation, and in several cases no other interpretation is possible.162 In no way, then, can the formula etasmdd eva be taken as evidence that the forms in -teh and -eh are ablatives. Turning to the commentaries, we find support for this conclusion. This support is not entirely unambiguous, though in the vast majority of instances it is clear that the commentators take etasmdd eva to mean 'for the very same reason'. Durga, in fact, explicitly expands the set expression to etasmdd eva kdrandt, 'for this very reason' in two instances.163 Several times164 the commentators add the word dhdtoh or a form in -teh after etasmdt, but more than once they themselves make it clear that they consider these additions to be genitive and not ablative forms.165
157 159 158 The context is dnasah RV 4 30 10 2 6, 5 1, 6 17b, 9 8, and 11 47 1.20a, 1.20b, 2 16 2.13, 2.17, 2.22, 4 13, 4.16, 4 18, 4 19, 4.21, 4.27, 5 2, 5.4, 5 14, 6.1, 6 17a, 8, and 10 44 ° 2.17, 4 16, 4 27, 5 2, and 6 17a 161 2.5, 3.9, 3.10, 4.5, 4.8, 5 1635 22, 5 26, 6.1, and 11.39 7, 162 E g., 4 19,4.27, 5.2 Durga at 5 2, 5 14 14 2 17, 2 22, 3 9, 3 10, 4 5, 4.8, 4 13, 4 16, 5 7, 5 22, and 5 26 6 165 Durga at 3.10, 6.1, 11.39, Skanda-Mahesvara at 3.9, 5.7, 5.26.

752

Indian semantic analysis

More problematic are the following statements. At 2.5 Durga states (D 1:179,8-9): etasmad eva dhdtudvaydd gamer gdter vd which it is most natural to interpret as ' etasmad eva, that is to say, from these two roots, namely from Vgam or Vga'. Skanda-Mahesvara (SM 11:43,1) too add dhdtudvaydt after etasmad eva. At 4.5 they remark (SM 11:208,17): tathd cdyam muso 'py etasmad eva dhdtos tathoktah. It is perfectly reasonable to interpret this as: 'and thus also this [word] musah is in that way stated as from this very [same] root'. At 7.29 (SM 111:110,3) they add dhdtupratyayasamuddydt after etasmad eva. This could very well mean: '[that is to say,] from the [same] combination of roots and suffixes'. If etasmad eva does not mean 'for the very same reason' it would have to refer to a verbal root and hence to a grammatical derivation. Apart from the fact that such an interpretation does not work in many instances, there is also the circumstance that etasmdt quite frequently is followed by an additional explanation, clearly causal.166 Even if etasmdt were to refer to a verbal root, this shows that the grammatical derivation would in any case be based on a causal link between the name and an action considered the cause of the name. Conversely, to conclude that etasmdt indicates a cause is not to say that in providing a causal explanation the etasmad eva formula does not also reveal a grammatical analysis which ultimately involves a verbal root; on the contrary, it does this in the same way as an analysis of the -andt type. This is "why boundaries are floating and why there are some instances in the commentaries where etasmdt may refer to a verbal root. But it is the causal link which is primary; that is to say, etasmdt does not refer to a form in -teh but states the reason why there is a relation between a particular word and the activity expressed by a verbal root. Signification and the genitive case: ity apy asya bhavati j

I shall now turn to material of importance for the possible interpretation of forms in -teh as genitives and for the usage of the genitive case in the Nirukta. It is the genitive case which is used to mark the relation between a word and that which it signifies, that is to say, the genitive is used to state that a word is a word for a particular thing. This is the case both in the Nirukta and elsewhere. I shall adduce a couple of examples: gaur itiprthivyd ndmadheyam, 'gauh is a name for the Earth' (Nir 2.5). Patanjali, in his Mahdbhdsya, uses the same genitive construction (Mbh 11:205,21): apradhdnasyopasarjanam iti samjnd kriyate, 'the technical name upasarjana is made for apradhdna "subsidiary word'".167
166 167

2 8, 2.22, 3.9, 3.10, 4 13, 4 16, 4.18, 4.27, 5 4, 5 26, 10.44, 11.39, and, for example, Durga at 45 The technical term upasarjana is defined A 1.2.43 prathamdnirdistam samdsa upasarjana and A 1 2.44 ekavibhakti cdpurvanipdte, the problem of whether the two terms are really synonymous need not concern us here.

The universe of Yaska

153

Now, when it is said that 'gauh is a name for prthivV this seems to me to mean both that gauh is a name for the Earth, that is to say, that it signifies the Earth, and that the words gauh andp/t/i/vfare synonyms. In fact, the first three chapters of the Nighantu rest on this assumption. Each section contains a group of synonyms the meaning of which is indicated by the adhikara-word of that particular section. In the statement gaur itiprthivyd ndmadheyam, then, we can observe a certain confusion between a term and that which it signifies. We can also observe that the genitive in this construction fulfils a double function. On the one hand, it is used to mark a relation between a word and that which it signifies. On the other hand, it is used to mark a relation between one linguistic element and another, that is to say, between two words. More specifically, the latter relation implies that one can substitute 'gauh' for 'prthivV. Indicating a word as a synonym of another is tantamount to saying that one word occurs in the meaning of another word. Disregarding the confusion between a word and the thing it signifies, one can argue that if a genitive case ending is used in this manner, it is in effect a substitutional genitive, what the Indian grammarians refer to as a sthdnasasthT. Definitions of synonymy are hard to come by in works on theoretical linguistics and philosophy of language, but the few we come by seem to presuppose or involve substitution criteria. Addressing the question of synonymy within one and the same language, Quine (1961:56) remarks: So-called substitution criteria, or conditions of interchangeabihty, have in one form or another played central roles in modern grammar. For the synonymy problem of semantics such an approach seems more obvious still. However, the notion of the interchangeability of two linguistic forms makes sense only in so far as answers are provided to these two questions: (a) In just what sorts of contextual position, if not in all, are the two forms to be interchangeable? (b) The forms are to be interchangeable salvo quo) Supplanting one form by another in any context changes something, namely, form at least; and (b) asks what feature the interchange is to leave invariant. Alternative answers to (a) and (b) give alternative notions of interchangeability, some suited to defining grammatical correspondences and others, conceivably, to defining synonymy. When it is stated in the Nighantu that two words are synonymous we do not have to worry about the problem of context because they are clearly stated to ibe synonyms in at least one context. The same holds good when Yaska states \gaur itiprthivyd ndmadheyam. That 'gauh9 and 'prthivV are synonyms, that is fto say, that they are substitutionable, entails that the name 'gauh' can occur in pthe place of, that is to say, in the meaning of, the name 'prthivV. \ With this background, I shall now turn to a set wording or formula, repeatedly employed by Yaska and the commentators alike, which is of considerable interest when it comes to understanding the use of this genitive and how, at lleast at one level, it relates to the genitive used in the analysis of the -teh type. ;fThe formula x ity apy asya bhavati is used when a nirvacana has just been given and a following word is analysed in the same manner. Roughly, one may

154

Indian semantic analysis

translate this as 'x is also of it', but a more detailed investigation of the instances occurring in the Nirukta itself seems called for. (1.17) tvisitom jvalitahl tvisir ity apy asya diptindma bhavatil > Sarup translates (1921:18): 'Tvisitah means shining. Of this word (the part) tvisih is a synonym of light.' I do not think Sarup has understood this correctly. In the first place he does not account for the particle apj, which is rather crucial to this set expression. I suggest the construction be interpreted as follows. First tvisitah is explained by the synonym jvalitah 'lighted; blazing; burning'; then, since Yaska remembers there is also a similar word tvisih, he adds that this too is a synonym of jvalitah, but because tvisih can mean 'vehemence' as well as 'light, blaze', he specifies the condition that this holds good only when it is a name for diptih 'light'. The passage could then be translated: 'tvisih is also [a name/synonym] of this [when it is] a name for light'. Durga (D 1:138,10-11) quotes the text as above and adds: apathitam api diptindmasu na kevalam ydny eva pathitdny anydny apy iti, 'even that which is not listed [is included] among the names for light, [that is,] not exclusively those which are listed, but also others'. This is a bit odd, since there is no section in the Nighantu listing diptindmdni, names for light; nor does there exist, as far as I am aware, any other lexicographical work with such a list. Probably he is referring to the jvalitp ndmadheydni at Nighantu 1.17. Skanda-Mahesvara (SM 1:109,6-7) quote the passage verbatim and add: iti prdsangikam tvisisabdasya nirvacanam, 'thus an occasional nirvacana of the word tvisiH\ (2.15) dinndmdny uttardny astaul. . . / tatra kdsthd ity etad anekasydpi sattvasya [ndrna]169 bhavatil kdsthd diso bhavantil krdntvd sthitd bhavantil kdsthd upadiso bhavantil itaretaram krdntvd sthitd bhavantil ddityo 'pi kdsthocyatel krdntvd sthito bhavatil djyanto 'pi kdsthocyatel krdntvd sthito bhavatil dpo ypi kdsthd ucyanteI krdntvd sthitd bhavantiti sthdvardndm I The following eight are names for the quarters . . . Among these kdsthd is [a name] for several entities. [Thus,] kdsthdh are the quarters: striding across (V&ram), they are set {^sthd)\ kdsthdh are the intermediate quarters: having crossed each other, they are set; the sun too is called kdsthd: striding across, it has set; also the strip of a shooting range170 is called kdsthd: striding out, it comes to an end; the waters too are called kdsthd: striding out, they come to a halt, thus [it is a name/synonym] of stationary waters. Again, we have a genitive construction, in structure parallel to ity apy asya bhavati, which states that a particular word is a synonym for several other
168 170 169 The context is RV 10 84 2 = AV 4 31 2 Omitted by several MSS I follow Durga here, who (D I 312,16) glosses ajyantah by sarapathdntah 'the end of the course of an arrow', pw, MW, and Apte identify it as the goal of a race-course

The universe of Yaska

155

words, although the primary relationship here is that between a word and the thing signified. (6.21) rnjatih prasddhanakarmd I [. . .] 171 rjur ity apy asya bhavatil rjuniti no vdrunahl ity api nigamo bhavatil rnjatih denotes the activity of perfecting/embellishing;... rjuh 'straight' is also of this: 'with straight guidance [let] Varuna [lead] us' (RV 1.90.1); this too is a quotation from Veda. Both rnjatih and rjuniti occur at Nighantu 4.3. In the present case, asya in the set expression refers back to rnjatih; in other words, it is equivalent to rnjateh which accordingly has to be understood as a genitive form. This is also how Durga understands it (D 1:668,7): rjur ity apy asya/va rhjater bhavati. Skanda-Mahesvara state the same (SM 11:466,8). (6.22) kuruh krntateh I kruram ity apy asya bhavatil kuruh 'Kuru' is of krntatih (^Ikrt 'cut'); kruram 'cruel' is also of this. Even Sarup has to deviate from his standard translation here (1921:106): The word krura (cruel) belongs to the same root also.' Durga (D 1:675,14) glosses asya by krntateh which is thus to be considered a genitive form. It is also worthy of notice that after quoting the analysis kuruh krntateh, Durga (ibid.,13) adds: sa hi satrun krntati, 'for he cuts to pieces enemies'. This makes it clear that he thinks of the analysis krntateh as involving much more than a mere reference to a verbal root. (6.27) amhuro 'mhasvdn I amhuranam ity apy asya bhavatil krnvdnn amhuranad uriil ity api nigamo bhavati/ amhurah means distressed;172 amhuranam is also [a name/synonym] of this: 'making free from the distressed' (RV 1.105.17); this too is a quotation from Veda. The word amhurah occurs at Nighantu A3. This is a case parallel to Nirukta 1.17 tvisito jvalitahl tvisir ity apy asya diptindma bhavati, discussed above. First a word from the Nighantu is explained, then another word is quoted as a synonym of it, asya unambiguously a genitive form. Nevertheless, Sarup translates (1921:108): The word amhuranam is derived from the same root also.' Similarly, Skold (1926:184) remarks in a note to the word asya: 'We should expect etasmdd eva.' As we have seen, we should not expect etasmdd eva since no reason is given here. We are dealing with a manner of indicating that another word too is a name for the same thing, ultimately indicating synonymy by means of a form in the genitive case. Durga supports the reading, adding only that there is an occasion for offering an analysis also of amhuranam since there is usage also of this word (D 1:687,12), a claim he justifies by quoting RV
171

Several MSS insert RV 10 76 1 as a nigama here. Durga, however, says that Yaska does not m give a nigama here (D 1.667,15) That is, amhas-vat-, amhah is explained at Nir 4 25

156

Indian semantic analysis

1.105.17. This indicates that for Durga the formula ity apy asya bhavati is common usage simply not worth commenting upon. The assumption that we are dealing with synonymity is supported by the commentary pf SkandaMahesvara (SM 11:482,14-15): asyoddharane pradarsitavye I etatsqmdndrthasya sdrupyaprasangdd amhuranasabdasya tdvan nigamam darsayati krnvann amhurandd iti I When an example of it [= amhura] is to be given, he simply gives the Vedic quotation krnvann amhurandt for the word amhuranam which has the same meaning as it, because there is occasion [for an analysis of that which is] of similar form. (10.17) m seva iti sukhandma I sisyateh I vakdro ndmakaranahl antasthdntaropalingi I vibhdsitagunahl sivam ity apy asya bhavati I sevah is a name for happiness [and] is of sisyatih {^sis 'stir'); the syllable va is a noun-maker; [and the word] has a substitute174 [v] for the penultimate [s\\ [and] optionally it takes guna; sivam is also of this. Here it is more difficult to determine with absolute certainty whether asya refers back to sevah or to sisyateh. I am inclined to believe that it is to the latter. Durga is not clear about his view on this, although he clearly considers sisyateh to be a genitive form (D 11:981,13): sisir dhdtuh, vakdrah pratyayah sisa ity asya, 'the root is sisih (Vizs), the syllable va is the suffix, [so it is] of this [root] sisA (Dhp 1.718)'. Skanda-Mahesvara too consider sisyateh to be a genitive form (SM IV:20,4). To the phrase sivam ity apy asya bhavati Durga just adds (D 11:982,3) agunapakse, 'on the option that there is no guna strengthening', and Skanda-Mahesvara comment similarly (SM IV:20,5): vibhdsitagunatvdc ca sivo 'pi siddhah, 'and because guna strengthening is optional here, [the word] sivah [sic] is also established'. (11.5) candras175 candateh I kdntikarmanah I candanam ity apy asya bhavati I candrah 'bright' is of candatih (^cand 'shine') which denotes the activity of being beautiful; candanam 'sandalwood' is also of this. The context is the explanation of the word candramdh from Nighantu 5.5. Sarup (1921:171), ignoring the genitive, translates: 'Candra "bright5' is derived from (the verb) cand, meaning to shine. The word candanam (sandalwood) is derived from the same root also.' Skold (1926:244), presupposing tliat we must be dealing with an ablative, remarks in a note to asya: 'We should have expected etasmdd eva? There can be no doubt, however, that asya is a genitive form and that it refers to candateh which therefore is a genitive form too. This is confirmed by Durga who remarks (D II: 1039,17-18): .
173 174

Cf. the previous discussion p. 117 above. The term upalingin occurs only here in the Nirukta, and I am not confident abput my trans175 lation. v.l candram eight MSS, candrdm occurs as a name for gold at Nigh 1 2.

The universe of Yaska

157

atha candrah kasmdtl candateh kantikarmanah/ nityakdnto hy asaul candanam lty apy asya/va candater bhavati / tad apt hi kdntam I Now, why candrahl 'It is of candatih (^cand "shine") which denotes the activity of being beautiful.' For it is perpetually beautiful. [The word] 'candanam is also of this' very same, [namely] of candatih, for that too is beautiful. Skanda-Mahesvara, on the other hand, do not seem to support the reading if the edition of the commentary is to be trusted here. First of all, they read candram and accept that candateh is the analysis if it is a word for gold as in Nighantu 1.2 (SM IV:57,5). They clearly consider candateh a genitive when they state (ibid.): candater dhdtoh kdntarthasya, 'of the root candatih which has the sense of being beautiful'. But they go on to remark (ibid.,6-7): candanasabdo176 'py ata eva dhdtoh prasangdd111 uktvd prakrtg,syaiva cam nirvacandntaram aha, 'having stated - because there is an occasion [for doing so] - that also the word candanam is from that very root, he states another nirvacana of the very [word] under discussion [=candramdh]\ I do not find the reading ata eva dhdtoh convincing because candateh has just been established as a genitive. It may of course be that Skanda-Mahesvara alternate freely between genitive and ablative interpretations. If the reading is genuine, then they would simply refer to the grammatical derivation of candanam from the root Acand, and their remark would no longer have any bearing upon the understanding of the expression ity apy asya bhavati. The material presented above is not vast, but I consider it of the utmost importance. It provides us with a set expression, a formula repeatedly used by Yaska; the commentators are quite familiar with it as well. All of the seven instances met with are illuminating. In two of the passages, tvisito jvalitah I tvisir ity apy asya diptindma bhavati (Nir 1.17) and amhuro 'mhasvdn/ amhuranam ity apy asya bhavati (Nir 6.27), the genitive asya in the formula ity apy asya bhavati is used to indicate that a word is used to signify the same thing as another word, that is to say, ultimately speaking that the two words involved are synonyms. So also in dinndmdny uttardny astau / . . . / tatra kdsthd ity etad anekasydpi sattvasya [ndma] bhavati (Nir 2.15). That is to say, in as much as one word is given in the sense of another word we are dealing with a substitutional genitive. The genitive indicates 'of this thing' as well as 'in the place of this word'. In two other passages, kuruh krntatehl kruram ity apy asya bhavati (Nir 6.22) and candras candateh I kantikarmanah I candanam ity apy asya bhavati (Nir 11.5), the genitive form asya refers to the preceding form in -teh. In the case of rhjatih prasddhanakarmdI... rjur ity apy asya bhavati (Nir 6.21) the genitive asya refers back to rhjatih which appears in the nominative case because this is how it occurs at Nighantu 4.3, and asya here is glossed rnjateh by Durga. In the remaining instance, seva iti sukhandmal sisyatehl. . . sivam
176 177

Corrected by Sarup (1934 623) from candasabdo0. 178 The editor remarks: 'pro1 nasti B C The editor remarks 'ca' nasti B C

158

Indian semantic analysis

ity apy asya bhavati (Nir 10.17), it is even hard to determine whether asya refers back to sisyateh or whether it indicates sivam as a synonym of sevah. What this information tells us, then, is that the same genitive form asya in the formula ity apy asya bhavati is used both to indicate that one word signifies the same thing as another word, thus ultimately indicating that the two words are synonyms, and to indicate a form in -teh, that is to say, a grammatical derivation. In some instances it is even hard to determine which alternative we are facing. Thus, the same formula ity apy asya bhavati refers both to a form in -teh and to the thing signified by a word. This cannot be coincidence. Ultimately the two alternatives work in the same way if we think in terms of substitution. When the genitive indicates that one word signifies the same thing as another word it means that one word is a synonym for the other. The words are thus substitutionable. And I take this formula to be indicative of how even the genitive in an analysis of the -teh type ultimately is to be understood, i^amely, as a substitutional genitive tantamount to the one used in indicating a synonym. For example, the analysis candras candateh (Nir 11.5) clearly indicates that candrah 'bright' is to be derived grammatically from candatih {^cand 'shine') but also that it is a synonym of the expression yo candati 'that which shines'. I shall substantiate this claim by turning to the iti satah type of analysis. The full version: analyses in -iti satah The analysis megho mehatfti satah actually occurs at Nirukta 2.21. It represents the most elaborate type of analysis met with in the Nirukta. In this case a finite form of the verb (mehati 'it rains') is incorporated into a nominal construction by means of the words iti satah. Here we have, I suggest, the clue to how the mechanism of a nirvacana analysis ultimately operates. ' Not more than twenty-five words are analysed in this way in the whole Nirukta, but they display some remarkable features: all of them Ire found in the Nighantu, eight of them are adhikdra-words from Nighantu 1-3, and five are names of deities. This indicates that we are dealing with an elaborate type of analysis employed only in very important instances. Accordingly, what this formulaic expression ultimately reveals will be of crucial importance for how we are to understand nirvacanas in general. I have dealt with all the instances of this type of analysis elsewhere179 and shall confine myself here to presenting some highly interesting facts and conclusions disclosed by a complete^ investigation: (1) At first sight an analysis of the iti satah type seems to present the same ambiguity with regard to case form as the analyses in -teh and -eh, but th$ crucial instance grhdh kasmdtl grhnantiti satdm (Nir 3.13) admits of no doubt:! the analysis is expressed in the genitive case. It is therefore evident that the sin-'
179

See Kahrs 1980 130-48, 233-50, see also 1984 150-2

The universe of Yaska

159

gular forms satah and satydh are genitives as well. This instance also reveals that sat-, the present participle of the root ^las 'to be', agrees in number with the word which is analysed. It takes the unique instance of aplurale tantum to reveal this important piece of information; a word is normally analysed in the singular number. (2) That sat- follows not only the number but also the gender of the noun being explained is clear from the following instances: usdh kasmatl ucchatiti satydh (Nir 2.18) and vdslti vdnndmal vdsyata iti satydh (Nir 4.16). It is evident from this that the relation sustained between a term and its corresponding explanatory expression has to be conceived of as being far more than just a reference to a verbal root. One immediate intention with an analysis of this kind is obvious: it enables Yaska to delimit the number of possible kdraka interpretations by embedding a kdraka analysis within the framework of an analysis which is formulated in highly technical language also in every other respect. The following examples may serve to illustrate this. In karma kasmatl kriyata iti satah (Nir 3.1) the word karman 'action' is connected with Vfer 'do; make', but in the passive, an analysis as karmasddhana. In mrtyur mdrayatiti satah (Nir 11.6) the word mrtyu 'Death' is connected with Amr 'die', but causatively: mdrayati 'it causes to die', an analysis as hetusddhana. Mere reference to the verbal roots and suffixes involved would not necessarily bring to light the full semantic content of these words, although both of them lend themselves to complete grammatical derivation as laid out in the Unddisutras. 18° The Unddisutras seem to derive mrtyu in the sense of marana 'death; dying', that is, as bhdvasddhana.m Yaska, on the other hand, explains mrtyu in the sense of 'Death', and only an interpretation as hetusddhana (mdrayati) would disclose the function of the god of death: he causes beings to die. The commentators leave us in no doubt that they consider a restriction with regard to kdraka interpretation the main intention behind an analysis of the iti satah type. For example, commenting on the analysis dyur ity ahno ndmadheyamml dyotata iti satah (Nir 1.6), Durga remarks (D 1:77,10-11): dyotata iti satah / dyotate tad iti dyur iti kartrkdrakamml sad iti yatra bruydt tatroccdnta eva kdrakaniyamo drastavyah I anyatra yathestham yojyam I 'dyotata iti satah': in as much as one says 'it184 is bright', [it is called] dyuh 'day' C [and] thus [to be interpreted as] kartr-kdraka. Wherever he (= Yaska) may add sat-, there will be seen the kdraka restriction to have m fact been stated. Otherwise [i.e. when sat- is missing] one must construe according to preference. J 0 p p u 4 144 a n d 3 2i5 DPU 6.73 and 1.135; cf A 3 3 1 unddayo bahulam, '[the suffixes] un etc 8 [occur] promiscuously'. 181 Yudhisthira Mimamsaka (1943 89, note 6) lists the v 1 mrlyante jand aneneti, 'people die by means of it', which would give an interpretation as karanasddhana 182 dyuh is listed under aharndmdm at Nigh 1 9 ?3 With v 1. and Rajavade 1921 49,4 for dyotater dyur ml dyotata iti kartrkdrakam 184 tad - dyuh, attested also in the neuter according to MW.

160

Indian semantic analysis

Similarly, Skanda-Mahesvara have this to say about the same analysis (SM 1:67,7-10): dyur lty ahno namadheyam, tad dhi rasmisambandhdd dyotate, sata ity updttam kdrakapangrahdrtham, dyotata iti evam satah kartrkdrakel%s yuktasya satah kartrkdrake na kdrakdntara ity arthahl evam yamo yacchatlti sata186 ity dddv api sarvatra nirvacane^1 sata ity etad vydkhyeyaml ' "dyuh" is a name for day', for it is bright since it is related to the rays [of the sun]. [The element] satah is employed in order to include [in the analysis an interpretation with regard to] kdraka; [that is to say, it is a name] 'for something which is' in such a way (evam) 'that one says (iti): it is bright', [that is to say,] for something which is in such a way that it construes in the sense of the zgent-kdraka, that is to say, in the sense of the agent-kdraka and no other kdraka. In the same way this [element] satah is to be explained also in such instances as (Nir 10.19) yamo yacchatlti satah [and so also] in all nirvacanas. Thus, the commentators agree that the main purpose of an analysis of this kind is to make a restriction with regard to possible kdraka interpretations.188 Now, if this is all Yaska wanted to say, he could have adhered to an analysis of the type mehatiti meghah. I therefore doubt that the commentators have got the whole story here. One more thing which distinguishes this type of analysis from an analysis of the -teh or -eh types is that the genitive of satah etc. operates in a different way. In an analysis such as megho mehateh the genitive form mehateh establishes, at least in the first instance, a relation with the word-form meghah, not with that which is signified by the word. In an analysis such as megho mehatiti satah there is no such direct relation between the term and the explanatory expression. Let me now turn to the way in which Durga interprets an analysis of the iti satah type (D 1:226,4-5): aha - meghah kasmat/ ucyate - mehatiti sihcaty asaul evam kartari kdrake satah asyaitad abhidhdnam bhavati megha iti I He (Yaska) asks: 'why meghahV It is said: 'it rains', that is to say, it sheds water, [and so] this name occurs for that 'which exists' in such a way, that is to say, in the sense of agent as far as kdraka is concerned, [and hence] 'meghah'. The syntax is a little cumbersome here due to the commentatorial style, but the following points are clear. Durga too recognises that the relation here is not directly between two word-forms. Nevertheless his interpretation is very
185

186 188

Corrected SM IV 608 from karakena. The dental n would imply kdrake na, a reading that makes no sense. The na may have crept m under influence of the subsequent kdrake na in the same passage. Corrected SM IV 608 from the impossible evam yamayos cid id iti sata ity°\ we are clearly 187 dealing with a quotation from Nir 10.19 Corrected SM IV: 608 from nirvacanena. The commentators frequently repeat this view in a more abbreviated form, for example Durga at Nir 2 13,2 18,2 21,3 1,3.9,3.10,4.16,10.34; Skanda-Mahesvara at 3.1,3.8,3 13,4.16,10.3.

The universe of Yaska

161

similar to the way in which he interprets analyses m -teh and -eh, for example (D 1:30,2-3): samdnpurvasya mndter abhydsdrthasya karmani kdrake samdmndyah, 'in the sense of karman as far as kdraka is concerned, [the term] samdmndya is of Amnd which has the sense of repetition189 and is preceded by [the preverbs] sam and d\ In other words, the only difference Durga sees in the two types of analysis is that in the iti satah case the kdraka interpretation has been spelled out and the fact that the relation marked by the genitive case ending is not directly between the verbal element and the term being explained forces him to supply asyaitad abhidhdnam bhavati, 'this is a name of that' or 'this name is of that'. The discussion of the ity apy asya bhavati material revealed that the distinction between a word and that which it signifies is blurred in the Nirukta. Similarly, in the present instance, Durga does not seem to distinguish sharply between word and signified thing. It is evident, however, that in an analysis of the iti satah type the primary relation is not between one word and another, but between a word and that which it signifies. This is clear because mehatiti san is self-referring and thus it would be difficult for it to serve as an ideal substitute for meghah. Another feature which brings this out is the very fact that sat- agrees in gender and number with the word that is being analysed. This makes sense only if the reference is to the thing and not to the word which denotes it. It is equally evident, though, that the distinction between word and denoted thing is blurred in nirvacana analysis. This is the case even within the framework of analyses of the -teh type; for example: vaydh sdkhd vetehl vdtdyand bhavanti, 'vaydh, [meaning] branches, is of/from vetih (Vvf 'move'); they are such as move in the wind' (Nir 1.4). Names are analysed but in relation to the things they name. It is worthy of notice that nowhere do Durga or Skanda-Mahesvara comment on the fact that satah follows the term being explained in gender and number, nor do they attach any other significance to sat- than its function of determining which kdraka is involved. I think more is implied by sat- than just this. I also believe that the genitive is to be understood in a different way. Accordingly, I think it is possible to pursue matters further here. Let me first of all suggest the following simple interpretation of megho mehatiti satah: 'meghah' is of that which really exists (satah) so that one says (iti) [of it]: it rains (mehati). According to this interpretation, the analysis determines the reference of meghah by giving a definite description as in the nirvacanas of 'Bhairava', and it provides a relation with a verb or a verbal root. A straightforward rendering of the above interpretation would seem to be that the word meghah belongs to something, namely a cloud, of which one can say that it rains. But if names are introduced by a description, this would be either as a theory of the meaning of
189

Cf. Dhp 1 976 mna abhyase.

162

Indian semantic analysis

names or merely as a theory of their reference. Yaska has stated clearly that his search is for the meaning of names. Let me therefore assume that we are dealing with a theory of the meaning of names. It is possible, as Itshall later argue,190 to interpret the genitive as marking the relation of 'being in the place of and to conceive of this 'place' as the meaning of a word in the sense that megho mehatlti satah could be interpreted: 'meghaJf is in the meaning of that which really exists {satah) so that one says (iti) [of it]: it rains (mehati). Under this interpretation the genitive is taken as a substitutional genitive, in the same way as the genitive used to indicate that a word is used to signify the same thing as another word, that is to say, ultimately speaking that the two words involved are synonymous. As pointed out in the previous section, Yaska is familiar with and indeed employs such a usage. This fits well with the fact that a considerable number of the words analysed by an iti satah construction are adhikdra-words in Nighantu 1-3. The adhikdra-word is semantically substitutionable for all the other words in the group indicating as it does the meaning common to all of them. Yaska also employs a substitutional genitive to let one linguistic item replace another without a focus on semantics (Nir 12.7): nir ity esa sam ity etasya sthdne, 'the [preverb] nih [occurs here] in the place of the [preverb] sam\ Thieme (1958:46) remarks on this particular instance that 'he [= Yaska] does not think of a technical substitution such as taught in Panini's grammar and can be sure to be understood by any grammatical layman'. This is precisely the point. The Nirukta is in no way a grammar processing language, but Yaska takes the understanding of a substitutional genitive for granted. Other instances where Yaska employs a substitutional genitive are: yat tu kihcid dgneyam taj jdtavedasdndm sthdne yujyate, 'but anything addressed to Agni is-'acceptable in the place of something addressed to Jatavedas' m (Nir 7.20); yat tu kihcid bahudaivatam tad vaisvadevdndm sthdne yujyate, 'but anything addressed to many deities is acceptable in the place of something addressed to the Visvadevas' (Nir 12.40). In the above nirvacana, then, mehatlti san would be grammatically and semantically equivalent to meghah were it not for the self-reference problem. But to the extent that the distinction between signifier and signified is blurred, to the same extent it becomes less crucial whether we talk of the linguistic expression mehatlti san or that which it signifies. The expression rfiehatiti san must involve language for it to be a nirvacana. Moreover, when it is stated at Nirukta 2.5 that 'gauh is a name for prthivV this implies both that 7gauh signifies the Earth, and that the words gauh and prthivi are substitutionable linguistic elements. I consider the iti satah construction to be a parallel case. When you say x iti satah y, this means that y is a name for that which is so that
190

See pp. 248 ff. below.

m

See BD 2.128.

The universe of Yaska

163

one says 'he who x-es\ If we move from the ontological to the linguistic domain, we have to replace the adjective satah by a relative pronoun: yo mehati 'he who rains'. The elements yo mehati and meghah are both grammatically and semantically equivalent and hence entities which can replace each other. What immediately requires further investigation, however, is the usage of the element sat-. Let me first of all turn to a usage of sat- met with in the Mahabhasya, for example in the discussion of A 1.1.60 adarsanam lopah. This rule defines the most general name for grammatical elision, lopa, as adarsana 'non-seeing'.192 Patanjali begins his discussion with the following observation (Mbh 1:158,2-8): arthasya samjnd kartavyd sabdasya ma bhild itil itaretardsrayam ca bhavatil ketaretardsrayatd I sato 'darsanasya samjnayd bhavitavyam samjnayd cddarsanam bhdvyate tad etad itaretardsrayam bhavatil itaretardsrayam ca naprakalpyanteI lopasamjhdydm arthasator uktamll 1 // kim uktaml arthasya tdvad uktaml itikarano 'rthanirdesdrtha itil sato 'py uktaml siddham tu nityasabdatvdd itilnitydh sabddh I nityesu ca sabdesu sato 'darsanasya samjnd kriyate na samjnddarsanam bhdvyate II It has been stated that a technical term should be made for the meaning and should not be established for the linguistic form. Moreover, there is mutual dependence involved. What mutual dependence? A name must be made for non-seeing that exists, and non-seeing is established by means of the term. This is something which involves mutual dependence, and something which involves mutual dependence is not acceptable [in grammatical operations]. It has been stated that the term lopa applies to meaning and to something that exists.' vt. 1 What has been stated? With regard to meaning this much has been stated: Hti has been used to indicate the meaning'.193 It has also been stated with regard to something that exists: 'but it is effected because linguistic forms are eternally established'.194 Linguistic forms are eternally established. And because linguistic forms are eternally established, a technical term is made for non-seeing that exists; nonseeing is not established by means of the term. ; The first issue to be raised here pertains to the claim that a technical term in grammar should apply to the meaning of the linguistic form and not to that linguistic form itself. That is to say, the term lopa refers to the meaning of r 'non-seeing' and not to the linguistic form adarsana. Patanjali makes a point 'of this because of rule A 1.1.68 svam rupam sabdasydsabdasamjiid which .teaches that a linguistic element (sabda) denotes its own form except in the ^case of a technical term in grammar. For a grammarian, who speaks about lanfguage, this is a distinction that simply has to be made in order to avoid the ^problem resembling the one which is solved by the modern use of quotation
492

Rule 1.1.60 and Patanjah's remarks on it have been discussed in detail by J.W. Benson m m 1990:124-40. A 1.1.44, vt. 3. A 1.1.1, vt. 9.

164

Indian semantic analysis

marks. The necessity of making a clear distinction between a* sign and its meaning is not mere pedantry, as Gottlob Frege was careful to point out (1893:4): Man wird sich vielleicht uber den haufigen Gebrauch des Anfuhrungszeichens wundern; ich unterschiede damit die Falle, wo ich vom Zeichen selbst spreche, von denen, wo ich von seiner Bedeutung spreche. So pedantisch dies auch erscheinen mag, ich halte es doch fiir notwendig. Es ist merkwurdig, wie eine ungenaue Redeoder Schreibweise, die urspriinglich vielleicht nur aus Bequemlichkeit und der Kiirze halber, aber mit vollem Bewusstsein ihrer Ungenauigkeit gebraucht wurde, zuletzt das Denken verwirren kann, nachdem jenes Bewusstsein geschwunden ist. Hat man es doch fertig gebracht, die Zahlzeichen fiir die Zahlen, den Namen flir das Benannte, das blosse Hilfsmittel fiir den eigentlichen Gegenstand der Arithmetik zu halten.195 In nirvacana analysis there is, as we have seen, a certain scope for confusion in this respect. Do we speak about a word or that which is signified by a word? A nirvacana is an analysis of a word, but with regard to how it is related to the thing signified. This is also where the wide semantic range of the term artha enters the picture. If an analysis of the Hi satah type is interpreted as I have suggested, there is no room for such confusion. It is the term which is analysed and it is analysed by a definite description, still language, of that which it signifies. The relation between the term and its corresponding explanatory expression happens through that which the term signifies. The second issue raised by Patanjali pertains to the difficulty that arises if a grammatical term were to provide a means of establishing a word, that is, if the term lopa is applied to non-seeing which already exists and is simultaneously used to establish non-seeing. This creates a problem of mutual dependence, itaretardsrayatd. The first vdrttika on A 1.1.60 states that both of these issues have been dealt with elsewhere. On the basis of arguments advanced in thosecontexts, it is settled that the term lopa signifies the meaning of adarsana and that the term is applied to something which exists. Patanjali quotes vt> 3 itikarano 'rthanirdesdrthah on A 1.1.44 nd veti vibhdsd. This vdrttika states that iti is included in the rule to indicate that the term vibhdsd has the sense of 'or not' and does not apply to the form of the linguistic elements na vd. Resorting to the principle of anuvrtti, the recurrence of part of a rule in a subsequent rule,
195

'Perhaps one is perplexed by the frequent use of quotation marks; I use them to distinguish between cases where I am talking about the sign itself and cases where I am talking about its reference. This might seem pedantic but I think it is necessary. It is strange how an imprecise way of speaking or writing, which originally might have been used due to laziness and for the sake of brevity, but in full awareness of its imprecision, can ultimately confuse thought after the awareness has gone. After all one has managed to take the number signs as numbers, the name as the thing which is being named, that is to say, the mere auxiliary device as the subject of arithmetic' "

The universe of Yaska

165

Patanjali seems to argue that the word iti of A 1.1.44 is read into A 1.1.60, having thus established that it is non-seeing, the meaning of adarsana, which is termed lopa. The problem of mutual dependence involves a fundamental debate on the nature of language: while linguistic forms are eternally established (nitya), they are also considered the product of grammar (kdrya, lit. 'to be produced'). Patanjali quotes vt. 9 on A 1.1.1 vrddhir ddaic, the rule that assigns the name vrddhi to the sounds a, ai, and au. In the Mahdbhdsya discussion of this rule the objection is raised that while the name vrddhi is applied to a, ai, and au which exist, these vowels are to be established by the term vrddhi. The case is an exact parallel to the one at A 1.1.60. Since there is room for confusion at this point, I quote the passage in question (Mbh 1:40,20-1): satdm ddaicdm samjhayd bhavitavyam samjhayd cddaico bhdvyante tad itaretardsrayam bhavatil itaretardsraydni ca kdrydni na prakalpante I A name must be made for a, ai, and au which exist, and a, ai, and au are established by means of the term. This is something which involves mutual dependence, and grammatical operations which involve mutual dependence are not acceptable. The passage continues by adducing a couple of examples. The difficulty of mutual dependence is then answered by vt. 9 which states that linguistic forms are eternally established, nitya. Once again the passage is an exact parallel to the one at A 1.1.60 (Mbh 1:40,26-8): siddham tu nityasabdatvdtII 9II siddham etatl kathaml nityasabdatvdt I nitydh sabddhl nityesu sabdesu satdm ddaicdm sarnjhd kriyate na samjhayddaico bhdvyante I 'But it is effected because linguistic forms are eternally established.' vt. 9 It is effected. How? Because of the fact that linguistic forms are eternally established. [For] linguistic forms are eternally established, [and] in as much as linguistic forms are eternally established, a technical term is made for a, ai, and au which exist; a, ai, and au are not established by means of the term. On the basis of this argument, the technical name vrddhi cannot serve to establish a, ai, and au which exist already since they, like other linguistic forms, are eternally established. Likewise, the name lopa in A 1.1.60 is assigned to non-seeing which already exists, and does not serve to establish non-seeing. One may argue that a slight difficulty arises concerning whether it is the word-form adarsana that exists already or its meaning non-seeing, but in the case of a, ai, and au this problem does not arise since here we deal .with phonological terms implying that ia> signifies a etc. That is to say, they are signified by their own form according to A 1.1.68 svam rupam sabdasydsabdasamjhd cited above. This is because a etc. on a phonological level do not carry meanings of their own, but in the context of the claim that a technical term should be made with regard to the meaning of a linguistic element that is eternally established, the problem seems to vanish since the

166

Indian semantic analysis

relation between word and meaning is said to be eternally established as well.196 It is thus ascertained that the name lopa applies to the meaning of the term by which it is defined and thus is something that exists. The word-form adarsana is not at question at all; the name lopa is given to real non-seeing and not to the word-form adarsana. Similarly, I suggest, it would be possible to argue that in an analysis of the iti satah type, we face the analysis of terms which signify something which actually already exists such that it lends itself to a particular definite description. Certainly, the term does not establish that which it signifies. The thing signified really exists so that it can be described in a certain way, and so that the definite description of it is true. It remains to explain why some words and not all words are analysed in this manner, particularly with regard to the element sat-. I have already given a brief rundown of the frequency of the iti satah type of analysis, arid the mere importance of the words analysed in this manner would to some extent explain why they are eligible for this more elaborate analysis. I have also tried to investigate the possibility of tracing a connection here to the ritual spHere.197 It is evident that several of the terms analysed by an iti satah construction, for example karman (Nir 3.1), megha (2.21), mrtyu (11.6), vaifas vdnndma (4.16), and certainly the deities, all have a relation to the sacrifice, that is to say, to a sacred, instantly efficacious ritual action. It is also difficult not to think of the innumerable satahlsatdm people who, through their life in truth, "are able to turn words into reality. It is on that basis tempting to paraphrase sat- as 'which exists in a way that is a true /real /efficacious existence' and conclude that it is used, for example, for karman as a name for 'ritual action' which is a truly real action as opposed to other actions which are not efficacious in the same way. I have tried to investigate this further in order to see whether it would be possible to track down an underlying system whereby terms specifically pertaining to ritual are analysed in this way, but I have failed to find anything conclusive. Indeed, it seems to me that such a system does not exist Moreover, several terms which pertain to the ritual sphere (e.g., chanddmsi and mantrdh Nir 7.12) are not analysed by means of an iti satah construction. Still, the ritual connection and the usage of sat- in connection with truth and making truth is worth keeping in mind since Yaska undoubtedly lived in a world of Vedic ritual, although perhaps no longer in that world of Vedic ritual which we come to know through the Brahmanas.
196

197

In the phrase siddhe sabddrthasambandhe, traditionally considered the first £art of the first varttika of the Mahdbhdsya. Patanjali (Mbh I 6,17) analyses the compound as* siddhe sabde 'rthe sambandhe ca, 'when [it is assumed] that the word-form, [its] meaning, and the relation [between them] are permanently established', it could alternatively be interpreted* 'when [it is assumed that] the relation between a word-form and [its] meaning is permanently established' Under both alternatives the relation between word-form and meaning is considered permanent This was first suggested to me by Harry Falk (private communication 29 6 1987).

The universe of Ydska

167

On the basis of this discussion, I suggest the usages of sat- in the Mahdbhdsya and in the ritual sphere be taken into account with extreme caution, and that in an analysis of the iti satah type sat- be interpreted in a way whereby it simply serves to sanction the definite description as true. In other words, the analysis megho mehatlti satah may be paraphrased: 'meghah is in the meaning of that which really exists (satah) such that it is true to say (iti) [of it]: it rains (mehati)\ With regard to an analysis of the type megho mehatlti satah, then, it has been established that satah is a genitive form and that sat- agrees in gender and number with the noun that is explained. This is because an expression such as mehatlti san refers to a real existent thing and not to a linguistic form. First of all, this makes it clear that the genitive here cannot be derivational, in the sense that 'x is a form of y\ Durga too recognises that the construction leaves no room for a direct relation between two linguistic elements. Supplying asyaitad abhidhdnam bhavati, 'this is a name of that' or 'this name is of that', he takes the genitive to mark a relation between a term and that which it signifies so that megho mehatlti satah is to be interpreted as 'meghah is a name for that which is such that one says: it rains'. Neither Durga nor Skanda-Mahesvara comment upon the fact that sat- agrees in gender and number with the noun that is being analysed, nor do they have much to say about sat-. The only thing they are eager to point out is that an iti satah analysis serves to indicate a specific kdraka interpretation, excluding any other kdraka interpretation. This nevertheless does not prevent them from offering a different one should circumstances so require (cf. SM IV: 102,1). Now, kdraka interpretation would easily have been taken care of also by a more simple analysis of the mehatlti meghah type. If kdraka were all Yaska wanted to specify, why would he bother to add sat- in the genitive case and have it agree in number and gender with the noun he explains? Whether we can say of any particular that it has necessary or contingent properties depends on the way it is described. In a nirvacana of the iti satah type, the element sat- serves to indicate a definite description which determines the reference of the term being analysed. Having also considered the usage of sat- in the ritual context and in the Mahdbhdsya, one may argue that it simply serves to sanction a definite description as true. An analysis of the iti satah type would then convey the following information: there is (sat-) an x such that it is true (sat-) to say of it 'it is Jt-ing' or 'it is being x-ed' etc., and therefore it is called V . To conclude, under this interpretation an analysis such as megho mehatlti satah may be paraphrased: 'meghah' occurs in the meaning of that which really exists such that it is true to say of it: it rains. In this way nothing has to be supplied; the interpretation simply rests on the way the genitive is to be understood, an issue to which I shall return in much detail below. In vydkarana you talk not of things but of how they are spoken about, and the approach is primarily derivational. Still, there is an overlap between meaning and form. For example, you would not add a suffix unless there is

168

Indian semantic analysis

meaning involved. The nirukta analysis is also concerned with words, but words in relation to things. This is why one finds causal and derivational explanations side by side. An analysis of the iti satah type provides a vast amount of information: it sanctions a definite description as true and presents sufficient linguistic material to yield the semantically and linguistically ideal expression for the word that is being explained. Note that the interpretation I have suggested for an analysis of the iti satah type is of a semantic nature. It brings out the artha of the word analysed, and this is exactly what the nirvacana method tries to achieve. This is also what the various types vof analysis have in common. The relationship between the various types of analysis It is now possible to see how the various types of analysis relate to each other. It is evident that the complexities of Yaska's technical language presuppose a long development. If we are to gain as full a picture as possible, we have to remember that we are dealing with elliptical ways of expression which have to be cautiously treated not only in relation to the metalanguage of Yaska which has its own historical presuppositions, but also in relation to the possibilities of interpretation and subtle conceptual distinctions developed by the later tradition. There is nothing which says that the commentators interpreted Yaska in an infallible manner. On the other hand, they themselves put their mark on a tradition which throughout employed and modified the methods exhibited by Yaska. As far as I know, there is only one significant exception: the iti satah type of analysis seems to be uniquely characteristic of the Nirukta. The analysis exemplified as megho mehandt was clearly well established before Yaska's time and frequently employed in later literature:'A further development of usages met with in Vedic literature, it is primarily a causal explanation indicating the action or event which is considered the cause of a name. Although it therefore establishes a relation between a name and a verbal noun and hence ultimately with a verb or a verbal root, it is evident that constructions of the -andt type can refer exclusively to the semantic content of a word. This explains why such constructions occur as additional semantic elucidations to analyses of the -teh type, while the reverse never is the case. Nowhere is an analysis of the -andt type seen to replace any of the other types. This holds good also for the formulaic expression etasmdd eva 'for the very same reason' which also indicates the cause for a name. A causal explanation is also provided by an analysis such as mehatiti meghah which in addition provides a kdraka interpretation which, as in the preceding instance, can be seen as a further development of usages met with in Vedic literature. Such kdraka analysis certainly establishes a relation to a finite verb and thus to a verbal root. Still, the primary relationship between a term and its corresponding explanatory expression is also in this case of a

The universe of Yaska

169

causal nature: x is called V because of the jt-ing or because it is jc-ing, is being *-ed, etc. As is well known, kdraka analysis permeates Sanskrit literature. While the two previous methods have a clear basis in Vedic texts, the remaining three are, as far as I know, confined to the Nirukta and later literature. They are technically more specialised and complex, and have one feature in common: they all exhibit a case form which is not immediately clear. In the case of the analyses in -teh and -eh it is evident from the previous investigation that it is not possible to rule out either a genitive or an ablative interpretation on the basis of possible attributes in the Nirukta itself. Although the commentators to an overwhelming extent treat them as genitive forms, the material from the Nirukta forces us to consider and indeed accept that the forms in -teh and -eh can be genitives as well as ablatives. In fact, both interpretations may be simultaneously possible. Let me first turn to the ablative interpretation. Under this alternative an analysis exemplified as megho mehateh or megho miheh could be seen as a purely derivational analysis, simply informing us that the word meghah is derived grammatically - and not etymologically or historically - from the verbal root mih. This more or less rules out a causal explanation, except as indicating a verbal root so to speak as a causa materialis, and could explain why analyses in -teh sometimes are followed by a verbal noun in -andt which clearly adds a causal explanation. It is clear that Yaska understood and indeed employed analysis into verbal roots and suffixes as well as secondary formation with taddhita suffixes.198 He does not employ the vaiydkarana term pratyaya 'suffix' but the term ndmakarana, literally 'noun-maker', implying that a suffix makes a noun, that is to say, derives a nominal from a verbal idea. Turning to the genitive interpretation of forms in -teh and -eh the picture gets more complex. It is quite clear that the commentators frequently take the genitive to establish a relation with a verb or a verbal root. First of all this fits well with their frequent addition of °arthasya, also added as glosses of Yaska's °karmanah. For example, commenting upon daksind daksateh samardhayatikarmanah, 'daksind is of daksatih (ydaks) which denotes the activity of rendering complete' (Nir 1.7), Durga adds the gloss samardhayatyarthasya, clearly a bahuvnhi compound in the sense of '[of daksatih] which has the sense of rendering complete'. There are also instances where the commentators are quite explicitly linking a form to a verbal root by the formula asyedam rupam, 'this form is of this . . . ' or similar. I shall adduce a few examples. On usiram (the name of a particularly fragrant root) Skanda-Mahesvara remark
198

For example, Nir 4 2 pdsyd pasasamuhah; cf. A 4.2.37 tasya samuhah which teaches that the taddhita suffix aN is added in the sense of 'collection of it' and A 4.2.29 pdsddibhyo yah which teaches that the suffix ya is added after the stems pdsa- etc. (gana 141: pdsddayah). This is no reliable argument for settling the question of chronology - Panini and Yaska may have based themselves on traditions commonly available to both of them - but it serves to show that Yaska was familiar with taddhita analysis according to the same linguistic principles that Panini employed.

170

Indian semantic analysis

\

(SM 11:45,5-6): vaseh Iraki krtasamprasdranasyedam rupaml kdmyate hi tat, sugandhitatvdt, 'this form is of vasih (Vvai 'desire; like') which has been subject to vocalisation of the semivowel when the suffix TraK (-Tra) is added; for it is desired because of its fragrance'. The genitive establishes a relation to a verbal root and the semantic reason for this choice of root is added separately. Similarly (SM 1:104,14—15): evam atrdvasdyeti na syater avapurvasya lyabantasya rupam, dhdtupasargayor lyapas cdtrdrthdsambhavdt, 'so, in this case avasdya is not a form of syatih preceded by the preverb ova and ending in [the absolutive suffix] LyaP (-yd), since the sense of root and preverb and LyaP is impossible here'. Likewise (SM 111:89,1-2): [dasyur] dasyateh 'tasu upaksaye dasu co* ity asya ksaydrthe vartamdnasyeti sesah, '[dasyur is of] dasyateh, i.e. of this [about which it is stated Dhp 4.106-7:] "[the root] tas [occurs] when wasting [is to be denoted], so also [the root] das"', it is of this \yhich occurs when there is the sense of destruction, thus the statement is to be completed'. Under these interpretations there would not be much of a difference between the genitive and the ablative alternatives. The genitive would establish a certain noun as a form (rupa) of a certain verbal root, and the ablative would similarly indicate a grammatical derivation from a verbal root. Determining what suffix is involved and which kdraka is left to the judgment of the reader. In fact, interpreted in this manner, the -teh and -eh forms of analysis would not convey more information than an analysis in -andt and less than the mehatiti meghah type. My hypothesis is that they are compressed ways of expression, eventually lending themselves to more elaborate interpretation. Mere indication of a verbal root is not what characterises the semantic quest of nirvacana analysis, and although this as an argument would be far from compelling, it is reasonable to assume, with the commentators, that we would be entitled to read considerably more into an analysis which simply indicates the verbal root by which a grammatical derivation would take place. The texts support such an assumption. In the Nirukta itself Several of the analyses in -teh and -eh are introduced by the word kasmdt 'why?' just like other types of analyses, including the less technical ones.199 This is also the case with the commentators, for example Durga (D 11:981,12-13): seva iti sukhandmal tat kasmdt I sisyateh, 'sevah is a name for happiness; why is that? sisyateK. Similarly, with a form in -eh (D 11:994, 16-17): atha garbhah kasmdt/ grbheh dhdtoh grndtyarthe vartamdnasya, 'now, why garbhah "womb"? It is of the verbal root grbhih (^grbh) which occurs in the sense of praising.' This not only allows for, but indeed necessitates reading a causal explanation into analyses in -teh and -eh even though, strictly speaking, all they do is indicate a verb or a verbal root. One could, of course, argue that in these instances kasmdt is not to be interpreted as 'why?' but as 'from what [verbal root]?', 200 but if this were the case it would show how one and the same
199

For example Nir 2 10 (hiranyam), 2 15 (disah), 2 18 (ratnh), 2 23 (vak), 3.9 (annam, balam) 200 etc. But see Kahrs 1983.

The universe of Yaska

171

formulaic expression can be interpreted in different ways according to the angle one may wish to approach a nirvacana from - semantic or derivational. In other words, if the Nirukta itself and the commentaries not only allow for but actually do read more information into analyses in -teh and -eh than what is actually stated, it is possible to argue that these analyses lend themselves to a fuller interpretation than what they formally allow for. This may include the genitive form itself. Although it is not possible to rule out the ablative interpretation of forms in -teh and -eh, it is an indisputable fact that the genitive interpretation gets fuller support from the Nirukta itself, and overwhelming support from the commentaries. This may be no coincidence. The same genitive form asya in the formula ity apy asya bhavati is used to indicate two different things. In some instances it informs us that one word signifies the same thing as another word, thus ultimately indicating that the two words are synonyms. In other instances it serves to indicate a form in -teh, that is to say, a grammatical derivation. Considering that in some instances it can even be hard to determine which alternative we are facing, it would certainly be possible to encode the notion of a substitutional genitive into the forms in -teh and -eh as well. Even in the technical language of the AstddhydyT, one and the same genitive is to be interpreted in different ways. For example, the genitive of A 6.4.1 angasya is to be interpreted as a sthdnasasthT, a substitutional genitive, in A 7.2.102 tyadddindm ah, but as an avayavasasthi, a partitive genitive, when it recurs by anuvrtti in A 7.2.106 tadoh sah sdv anantyayoh. Let me now finally turn to the way in which the interpretation of an iti satah type of analysis outlined above relates to an analysis in -teh and thus ultimately to one in -eh. In Nirukta 12.5 Yaska deals with the name Usas as the name of a goddess belonging to the heavenly sphere (dyusthdnd devatd), analysing it by usd vasteh kdntikarmanah. This pertains to the word usah listed at Nighantu 5.6. The same passage then gives an analysis of her name as a goddess belonging to the middle region: ucchater Hard mddhyamikd. This pertains to the word usah listed at Nighantu 5.5. In the context of Nighantu 5.5 Yaska's only remark on the name is (Nir 11.46): usd vydkhydtd, 'usdh has [already] been explained'. Now, this must necessarily refer to the one and only previous analysis, namely (Nir 2.18): usah kasmdtl ucchatiti satydh, 'why [the name] usdhl [Because] it occurs in the meaning of she who really exists such that it is true to say [of her]: she shines (ucchati)\ This analysis pertains to usah as the adhikdra-word of Nighantu 1.8 (sodasosondmdni). This warrants the following conclusion. Since Yaska at Nirukta 12.5 en passant gives the analysis ucchateh of the same usah that he said at Nirukta 11.46 had been explained earlier and which was analysed at Nirukta 2.18 by an iti satah construction, it is evident that he can substitute an analysis of the iti satah type with an analysis formulated by means of the artificial verb-noun in -teh. It is then possible to argue that ultimately the two kinds of analysis are to be understood in the same manner. We are dealing here with elliptical ways of

172

Indian semantic analysis

expression which lend themselves to more than one interpretation according to the different layers of Yaska's technical language, all of them interpretations which are simultaneously valid. The question also arises as to how the later tradition looks at the relationship between the iti satah and -teh types of analysis. Let me first turn to the commentaries on the analysis of usdh given at Nirukta 12.5. Since Durga and Skanda-Mahesvara are in close agreement here, suffice it to cite the remarks of Skanda-Mahesvara (SMIV: 101,14-102,1): usah kasmatl vasteh kantyarthasya I kdntd hy asaul ucchater vdl yathoktam 'usah kasmat ucchatlti satyah' iti dyusthdndyd201 vikalpahl madhyamasthdnd ucchater eva202/ tad aha ucchater itareftV sd hy udakdni vivdsayati vivdsyate vd meghdn madhyamenal Why 'usdh9? [Because it is] of vastih (^Ivas) which has the sense of being beautiful/splendid (Dhp 2.70); for she is beautiful/splendid. Or else \fusdh9 is] 'of ucchatih (^vas/uch203) "shine; expel"'; [that is to say,] just as it has been stated [Nir 2.18]: 'why usdh? [Because it occurs] in the meaning of she who really exists such that it is true to say [of her]: she shines/expels', it is [also here at Nir 12.5] an alternative for [usdh considered as] belonging to the heavenly sphere. [When, on the other hand, usdh is considered as] belonging to the middle sphere, it is only of ucchatih; therefore he says 'this other [usdh] is of ucchatih\ for she expels (vivdsayati) water or is [herself] expelled (vivdsyate)204 from the cloud by him who resides in the middle sphere.205 It is clear from this that the commentators too are happy to replace or refer to the analysis ucchatlti satyah with the analysis ucchatek Another example is the analysis of vratam 'action; vow' 206 (Nir 2.13).207 Durga comments on idam apitarad vratam etasmdd eva nivrttikarma as follows (D 1:207,15-208,3): vratam iti karmanama vrnotlti/ evam kartari kdrake satah vrnotehl tad dhi karma subham ca krtam sadd vrnoti kartdraml idam apitarad vratam etasmad eva nivrttikarma yamaniyamdkhyam bahirvedikam dntarvedikam vd yad agnisannidhdv upeyate, tat punah varayatlti evam satah tad visayesu pravartamdnam purusam stryddisu20* vdrayati tad anusrtya sa puruso nivartate 'vaklrnT bhavisydmity etasmdd bhaydtl
201 203

202 v.l. °sthdnd vd vi° MSS B and C. Durga adds vivdsandrthasya, genitive. There is a lot of confusion in lexica with regard to various roots vas/uch/us; cf. Dhp 1.231, 1.727, 6.14; see Kahrs 1980:237-41. , 204 This analysis as karmasddhana comes as a surprise since Yaska specified kartr by the ucchatlti satyah analysis at Nirukta 2.18. But as a nigama for usdh (mddhyamikd) oiNighan 5.5 Yaska offers RV 4.30.10 aposa dnasah sarat sdmpistdd aha bibhyusT. Geldner translates 'Die Usas lief von dem zerschlagenen Karren aus Furcht weg.' The commentators take ucchatih in the sense of 'expel; drive away' as the causative of virtvas, e.g. Durga 1:220,8 where Usas clearly expels darkness, or 1:368,3 where vivdsayati is glossed'ndsayati 'drives away; expels'. This makes the interpretation as karmasddhana completely straightforward. 205 That is to say, by Indra who conquers Usas in RV 4.30.10. 206 Nigh 2.1 lists vratam under karmandmdni. '207 See p. 134 above, and Kahrs 1980:234-6. 208 So read with Rajavade (1921) for purusastryddisu.

The universe of Ydska

173

'vratam is a name for action: it envelops'; [thus vratam] is [a name] 'of that [action] which really exists' in such a manner; that is, in the sense of agent as far as kdraka is concerned; that is to say, it is of vrnotih (Vvr 'cover; envelop'); for that action which has been done, even if it is good, always envelops the agent; this other vratam too, [i.e. vratam in the sense of] action of abstention [is so called] for the very same reason, that is to say, [action] known as yama and niyama (roughly: 'restraints and restrictions'), inside the sacrificial ground or outside it, that is, that [vratam] which takes place in the vicinity of the sacrificial fire, also that [vratam] 'causes to be enveloped'; that is to say, [it is a name] 'of that which really exists' in such a manner, in as much as it makes a man engaged in his activities cover himself up in front of women etc., and as he adheres to this [vratam] that man abstains out of fear from [having to make the statement:] 'I shall be one who has ejaculated.' Once more it is clear that an analysis of the iti satah type can be replaced by an analysis in -teh, which again is picked up by the expression etasmdd eva 'for the very same reason'. It is worthy of notice that Durga pays no heed to his earlier interpretation of the genitive satah as marking a relation with the thing signified, as the relation specified by vrnoteh clearly holds between the linguistic forms vratam and vrnotih. This little investigation makes it evident that we can construct a line from the analysis in -andt which gives a causal explanation, to the mehatiti meghah type which in addition offers a fully formulated kdraka analysis, to the iti satah type of analysis which, eventually, may be referred to by an analysis in -teh or -eh. There can be no doubt that the iti satah type of analysis is superior to any of the others. It incorporates all the information conveyed by each and every one of them, and, in fact, gives you more. By way of a definite description it gives a causal explanation of why a word is the right word for something, it offers a kdraka interpretation, it establishes a relation to a verb and hence to a verbal root, and it sanctions the definite description determining the reference as true. On the basis that Yaska employs a genitive to indicate a substitution procedure as well as to indicate that which is signified by a word and thus ultimately its synonym, that this same substitution principle underlies three of the five chapters of the Nighantu where words are arranged in groups of synonyms, I have argued above for an interpretation of the genitive in an iti satah type of analysis as a substitutional genitive or what the Sanskrit grammarians call a sthdnasasthu Under this interpretation of the genitive nothing has to be supplied. Now, derivatives of sat- are common words for truth in Sanskrit. What an iti satah analysis offers at the end of the day, then, is nothing less than a theory of truth as a theory of meaning all set up according to a substitutional model. Either one considers meghah the substitute for the ideal linguistic expression mehatiti son, disregarding in that case the problem of self-reference and the fact that the relation marked by the genitive strictly speaking is with the thing signified, or one considers the relation to be to the meaning of the

174

Indian semantic analysis

thing signified, the latter alternative requiring the construction of an ideal linguistic expression that involves a relative pronoun: yo mehatu The term meghah can replace the ideal expressions mehatiti son or yo mehati because they mean the same thing, one replacing the other at the relevant location in semantic space just like in the Paninian system some linguistic item may replace another in some specific context at the relevant location itt grammatical space. Now, as demonstrated above, it is evident that Yaska can replace an analysis of the iti satah type with an analysis formulated by means of the artificial verb-noun in -teh. It has also been established that the same genitive form asya in the formula ity apy asya bhavati is used to indicate that which is signified by a word, ultimately its synonym, as well as a form in -teh, and that in some instances it can even be hard to determine which alternative we are facing. This overlap illustrates that even a formulaic expression works in more than one way. It is then possible to argue that ultimately the two kinds of analysis are to be understood in the same manner, and that we are entitled to read more into an analysis in -teh or -eh than a mere reference to a verb or a verbal root. Note that I am not saying that the latter types of construction actually exhibit the same information as an iti satah construction, but that an investigation of the relationship between the various forms of analysis warrants the conclusion that they can be decoded to reveal the same information. Indeed, it is possible to argue that ultimately all nirvacanas are to be understood according to a substitutional model. An understanding of nirvacana analysis according to a model of substitution works well also for those analyses which are expressed in a less technical language, as for example the nirvacanas of 'Bhairava' discussed earlier. It also explains how one and the same word can be analysed in more than one way, and how seemingly exclusive explanations can be simultaneously valid. They are simply alternative placeholders in semantic space. It now remains to investigate the model of substitution and its standing in other parts of Sastric Sanskrit literature, and to see how the results of such an investigation fit in with the substitutional interpretation of nirvacana analysis.

5 Substitution

Whether or not it may licitly be considered a general pattern in Indian thought, the model of substitution is certainly a well-developed methodological procedure in Paninian grammar and in the ritual Sutras. And the science of grammar was certainly influential. Louis Renou went as far as to claim that 'adherer a la pensee indienne, c'est d' abord penser en grammairien'.l In simple words, the Indian grammarians adopted a model whereby stages in the linguistic derivational process are accounted for by saying CY occurs in the place of X' as opposed to 'X becomes Y \ The latter procedure was certainly known to the Indians and is met with in the Pratisakhya texts where changes of sounds are accounted for by letting one item change into another according to a morphophonemic technique where the items X and Y are denoted by a nominative and an accusative form respectively. Here the item Y is termed the vikdra 'modification; difference', while X is termed the vikann, literally 'possessing modification'. This procedure is clearly older than Panini2 although available evidence fails to ascertain that all Pratisakhya texts are earlier. The primary concern of the Pratisakhya texts was the relation between the Samhita or continuous text of the Vedas and the Padapatha, the analysed or word-for-word text. That is to say, they sought to describe those modifications or differences, vikdms, which distinguish the Samhitapatha from the Padapatha. Accordingly, the bulk of Pratisakhya rules pertains to changes of single sounds. The Paninian substitution procedure, on the other hand, involves not only single sounds but also larger units such as bases and suffixes. These units may even be theoretical terms such as the lakdras, the ten abstract verbal tense and mood suffixes which at the level of abstract syntax mediate between semantics and morphology.3 It is then no longer convenient or, indeed, possible to
1 3

'Being true to Indian thought is first of all to think as a grammarian'; Renou and Filliozat 1953, 2 § 1519 Cf Thieme 1958 45, Cardona 1969 32 f, 1976 274 f Rule A 3.4.69 lah karmani ca bhdve cdkarmakebhyah teaches that an L-suffix is added to a verbal root (A 3 1 91 dhatoh) to denote - in addition to the agent {kartan A 3 4.67) - the object or, m the case of intransitive verbs, the mere activity expressed by the verbal root Moreover, these L-suffixes serve to mark tenses. Other rules define what specific tense and mood are denoted by each particular L-suffix when they are introduced at particular stages of a grammatical derivation. For example, A 3 2.110 lun teaches that the aonst L-suffix luN is added after a

176

Indian semantic analysis

- .

talk about a vikara or modification, the substitution method being adopted and fully developed instead. The element which is to undergo substitution is indicated by the genitive case suffix, while the substitute is given in the nominative. A linguistic element has a possible domain of applicability which in Paninian terminology is referred to as the sthana, literally 'place; activity of standing', a term I shall return to in much detail further on. The substituend, the element which is to undergo the substitution operation, is called the sthanin, literally 'place-holder; placing-possessor', and the substitute, the element which replaces the sthanin, is called the ddesa, literally 'instruction; specification'. I shall discuss these two terms in turn. The sthanin The term sthanin has been studied in detail by A. Wezler (1972) who, with regard to Panini, renders it as (ibid.: 12): 'das, was durch eine ihm eigentiimliche / zukommende Stelle (d.h. eine Stelle, an der es erwartet wird (prasakta)4) charakterisiert ist'.5 He adds (ibid., note 30): 'Das ist die etymologische Bedeutung von sthanin- und nicht "ce qui [etait] a telle place [et n'y est plus]" ',6 thus condemning the rendering of Renou (1941-2:130). Wezler'slendering of sthanin is based on a conception of the secondary suffix -in advocated by Thieme (1955:191): 'Mit einer Formation auf -in- wird jemand oder etwas als der Besitzer (im weitesten Sinne) von einer ihm eigentumlichen Sache (die durch den dem Formans vorausgehenden Nominalstamm genannt oder bezeichnet ist) charakterisiert.'7 Although Thieme cautiously adds 'in the widest sense', I do not find this rendering very helpful, nor do I, consequently, accept Wezler's rendering of sthanin. I simply do not think it is possible to come up with such an exact and detailed mdaning for any secondary suffix in Sanskrit, be it on the basis of the observations of the indigenous grammarians or on the basis of Indo-European linguistics. Taking -in as a general possessive suffix, I prefer at this stage to render sihana as 'place' and sthanin as 'place-holder'. verbal root when the action refers to the general past (A 3 2 84 bhute). These abstract L-suffixes are subsequently replaced by finite verbal endings or by participial suffixes which may themselves be subject to substitution. The advantage of introducing the abstract level of L-suffixes m the grammatical process is that at this level all verbal endings can be said to signify agents, objects, or mere activity in relation to the action expressed and can thus be treated according to what they have in common at this level. This is Wezler's way of incorporating the indigenous interpretation ofsthdna asprasanga 'possible appearance, chance' and sthanin accordingly as prasakta, I shall return to the possible interpretations of sthdna further on. That which is characterised by its own peculiar/ due place (i.e. a place at which it is expected {prasakta))' 'That is the etymological meaning of sthanin- and not "that which [was] in such a place [and is no longer there]"' 'With a formation in -in someone or something is being characterised as the possessor (in the widest sense) of a thing which is peculiar to the possessor, (the thing is named or designated by the nominal stem which precedes the forming element).'

4

5 6 7

Substitution

177

As noted by Wezler (1972:13), Renou (1941-2:131) observed that Vedic ritual did not know the term sthdnin in any other sense than 'qui est a sa place' ('which is in its place') and adduced text-places to buttress his claim. Adding further evidence for this, Wezler (ibid.) finds it works also for the opposite concept asthdnin 'not in its place'. From this he moves on to asthdna which, as the basis for the derivation asthdnin, can be rendered 'unrechter Ort' ('improper place'), and sthdna accordingly as 'ein rechter, passender Ort' ('a proper, fitting place') although there do not seem to be any text-places in Vedic literature with this meaning. In other words, Wezler moves from asthdnin 'nicht am rechten Ort befindlich' ('not in the proper place'), to asthdna 'unrechter Ort' ('improper place') and, via sthdnin 'am rechten Ort befindlich' ('in the proper place'), to establish a sthdna 'ein rechter Ort' ('a proper place') and a locative sthdne 'am rechten Ort' ('in the proper place'). Wezler (ibid.: 14) then turns to instances where sthdne, 'einer Postposition vergleichbar' ('comparable to a postposition') is used in the sense of 'anstatt; anstelle' ('instead of; in place of). The text-places adduced for this claim are all reducible to the formula tasya sthdne. We have already encountered such a usage in the Nirukta (Nir 12.7): nir ity esa sam ity etasya sthdne, 'the [preverb] nih [occurs here] in the place of the [preverb] sam\ Indeed, this is normal Sanskrit usage at all stages of the language in the sense of 'in the place of it', and I see no reason why sthdne is to be considered a quasi-postposition. This conclusion is supported by the observations made by Wezler concerning the parallel instances where loke 'im Sinne von "an der richtigen, passenden Stelle" verwendet ist - so wie es fiir das semantisch sehr nahestehende sthdne in der Tat zu erschliefien ist' 8 (ibid.: 16), an interpretation which, notably, does not work with loke in a genitive construction parallel to tasya sthdne. Wezler admits that 'ein loke mit genitivischem Attribut = "an der/die Stelle, an die [eigentlich] x paBt/gehort" jedoch erweckt zumindest starke Zweifel'9 (ibid.: 17). He concludes, however, that sthdne preceded by a genitive attribute is equivalent to German '[an]statt, anstelle' in numerous text-places in the Kalpasutras, and as such is to be conceived of as a quasi-postpositional expression of pre-classical Sanskrit. Wezler (ibid.) transfers his interpretation of sthdne with a genitive attribute to the Astddhydyl, and interprets A 1.1.49 sasthi sthdneyogd, a rule to be discussed in much detail further on, as 'eine Sechste [Kasusendung] (d.h. ein Wort, das auf eine Genitivendung auslautet) ist [innerhalb der Astadhyayl] mit sthdne ("anstatt, anstelle") zu konstruiren',10 claiming that Panini's usage presupposes the usage of sthdne and that Panini resorts to ordinary Sanskrit usage.
8 9

. where loke 'is used in the sense of "at the right or fitting place" - as it is in fact to be inferred for the semantically close sthane\ 'A loke with a genitive attribute = "at the place at which x [actually] fits / belongs" evokes at least strong doubts' 10 'A sixth [case ending] (l e , a word that ends in a genitive ending) is [within the AstahyayT] to be constructed with sthdne ("instead of, in place of").'

178

Indian semantic analysis

Accordingly, not only rule A 1.1.49 but also the term sthanin has to be understood against the linguistic usage of the time as demonstrated by the Kalpasutras. But Wezler relies on his previous arguments and interpretations for his final conclusion (ibid.: 19-20): Denn wie aus sthanin-, 'am rechten Ort befindlich' - zusammen mit asthdnin-, seiner Kontradiktion, und asthdna-, 'unrechter Ort' - ein sthdne, 'am rechten Ort' zu erschlieBen ist, so darf man andererseits folgern, daB neben sthane, '[an]statt, anstelle', auch ein sthanin-, 'was durch eine ihm eigentumliche Stelle [namlich eine solche, an der es erwartet wird] charakterisiert ist', als allgemeine Bezeichnung fur die verschiedenen speziellen Begriffe, die jeweils als genitivisclie Attribute zu sthdne fungieren, bereits vor Panini existiert hat, daB also nicht erst Panini das in den Ritual-texten nur im Sinne von 'am rechten Ort befindlich' belegte sthaninauch in dem technischen Sinne von 'urspriingliches, durch ein Substitut zu ersetzendes Element' verwendet hat. Und wer die Zulassigkeit dieses Schlusses verneinen sollte, muB immerhin zugeben, daB der dann zuerst von Panini im Sinne von 'Substituendum' gebrauchte Terminus sthanin- nicht vorf dem sthdne, 'anstelle', der Kalpasutras getrennt werden kann.11 Indeed, I agree that it would be difficult to separate Panini';s usage of the term sthanin from the usage of sthdne with a genitive as met with for example in the Kalpasutras. But Wezler's aim is to take us to this contusion through the stages of his derivation - sthanin > asthdnin > asthdna > sthdna > sthdne a derivation proceeding in a rather unexpected order. Rich as his article is in observations, its interpretations and conclusions come across as unnecessarily cumbersome. To me it seems perfectly sufficient that we resort to the ordinary Sanskrit usage tasya sthdne interpreted simply as 'in the place of it' and sthanin as 'place-holder' to make the terminology and practice of the substitution procedure of the Astddhydyi emerge as a completely naturaPproduct of the J linguistic usage of its time. *A The Mesa The substitute, the element which is to replace the sthanin, is termed the ddesa. According to Thieme (1968), the term also refers to the substitution itself.12
11

12

'On the one hand one can make an inference from sthanin- "being in the proper place" together with its contradiction asthdnin-, and asthdna- "improper place" - to a sthdne "in the proper place". One may on the other hand infer that together with sthdne "instead of; in place of" also a sthanin- "that which is characterised by its own peculiar place [i.e,, a place at which it is expected]" has existed, even before Panini, as a general designation for the particular notions which act as genitive attributes to sthdne. That means that Panini Was not the first to use sthanin- which in the ritual texts had only been used in the sense of "being in the proper place", also in the technical sense of "original element which is to be replaced by a substitute". Those who want to deny that this conclusion is permissible have at least to admit that the term sthanin- which Panini uses in the sense of "substituend" cannot be separated from the sthdne "instead of" of the Kalpasutras.' G. Gren-Eklund (1984:118, note 53) suggests, in my view correctly, that Thieme's distinction between 'Substitut' and 'Substitution' may be due to our own languages and logic.

Substitution

179

Indeed, Thieme argues that adesa had the sense of 'Substitut; Substitution' already in the Brahmanas and oldest Upanisads (ibid.:262; my emphasis): Nun ist adesa nicht nur als 'Anweisung, Vorschrift', sondern auch als wissenschaftlicher Terminus mit einer Spezialbedeutung bei den Grammatikern belegt. Panini braucht adesa im Sinn von 'Substitution' (1.1.48, 7.2.86) und 'Substitut' (l.i.56,8.3.59,6.4.120,8.2.5). Patanjali hat auch das Verb a+dis 'substituieren', 'als Substitut lehren'. Es liegt nahe zu fragen, ob wir in der vedischen Prosa ebenfalls mit einem adesa 'Substitution, Substitut' rechnen dtirfen. Wir hatten damit, was wir suchen: die Benennung eines vom Lehrer gelehrten Verfahrens ('Substitution', 'Ersetzung') anstelle einer Bennung [sic] einer vom Lehrer gebrauchten Lehrmethode ('Vorschrift'). Zu untersuchen ware, ob das in den (alteren) Upanisads gelehrte Verfahren adaquat als 'Substitution', 'Ersetzung' benannt werden kann. Wir selbst wiirden wohl 'Identifizierung' sagen. 'Substitution' meint jedoch im Grund dasselbe, wenn auch in einer klein wenig anderen Beleuchtung. Wenn a-b ist, heisst das ja, daB man im Prinzip b fur a und a fur b 'substituieren', a und b gegenseitig 'ersetzen' kann. Vielleicht darf man sogar meinen, der Ausdruck 'Substitution' sei angemessener als 'Identifizierung'. Denn in den meisten Fallen handelt es sich gar nicht um eine gegenseitige 'Austauschbarkeit' - also eine vollstandige Identitat -, sondern nur darum, zu erkennen, daB fur den Wissenden a durch b zu ersetzen ist, wenn man es in seinem Wesen, seiner Essenz richtig fassen will, wahrend nichts damn liegt b durch a zu ersetzen.13 This passage is problematic. Although his reasoning is clearly circular, there can be no doubt that Thieme takes 'Identifizierung' to mean 'identity' since he introduces the equation mark. This is simply incorrect. Moreover, 'Identifizierung' ('identification') and 'Substitution' ('substitution') does not 'im Grund' ('in reality; really') mean the same thing. First of all, a and b are not names but variables. Once they are replaced by names it is clear that they are not so readily substitutionable even if they surround an equation mark. Secondly, only singular terms can stand on each side of an equation mark. If V and 'fc' are singular terms, 'a = V is true if and only if V and '&' have the same extension, that is, are co-referential. For example, 'the Earth = the third
13

'For the grammarians adesa does not denote just "order, rule" but it is also a scientific term with a special meaning. Panini uses adesa in the sense of "substitution" (1.1.48, 7.2.86) and "substitute" (1.1.56, 8.3.59, 6.4.120, 8.2.5). Patanjali has also the verb a+dis "to substitute", "to teach something as a substitute". This raises the question whether we can also expect an adesa "substitution, substitute" in Vedic prose. If so, we would have found what we were looking for: the naming of a procedure ("substitution", "replacement") which is taught by a teacher instead of a teaching method (rule) which is used by a teacher. It is to be examined ,-. whether the procedure which is being taught in the (older) Upanisads can be adequately named "substitution" or "replacement". We ourselves would probably say "identification''. In reality "substitution" means the same thing although it shows the phenomenon in a slightly different light. If a = b that means that in principle b can be "substituted" for a and a can be "substituted" for b, that is to say, a and b can be "replaced" mutually. Perhaps one can even say that the expression "substitution" is more adequate than "identification". For in most cases there is no mutual "replaceability" - that is to say, complete identity. What is important is to recognise that for one who knows a is replaceable by b if one wants to understand it correctly in its being or essence. It is not important to actually replace b by a.'

180

Indian semantic analysis,

planet from the Sun' is true. But this means that 'the Earth' and 'the third planet from the Sun' refer to one and the same thing, which would make identification easy but substitution very difficult except on a linguistic level. You cannot substitute something with itself. Nor is it correct, as Thieme himself recognises, to maintain that when a is identified as b then a can replace b and vice versa in all contexts. In an extensional language any two predicates which are true of the same objects are interchangeable salva veritate. Identity is a relation which relates a thing to itself and not to any other thing. If we have two classes which are to be filled it is a relation x=x or x=y. The statement 'x =x' is necessarily true. However, if we turn to the question of substitutionality we may get problems with the statement 'x=y\ Leibniz's law, the indiscernability of identicals, claims that if there is identity, then the objects have all predicates in common. That is to say, if there is a true statement of identity, the two terms are substitutionable in any true statement and the resulting statement will be true. This is not the case]. It resembles identity, but it is not identity. There may be differences in the objects at a given time in a given place. Thus contingent identity is different from necessary identity. It is necessary that a given object is identical with itself, For example, if we say: 'the Morning Star = the Evening Star', the object pointed out satisfies Leibniz's law. We cannot imagine that the one exists without the other. But in the statement: 'the statue = the piece of bronze' there is a relation which is not quite like the previous one, even if we call it a contingent identity. That identity does not necessarily allow for linguistic substitution will be clear from the following example. The statements 'Giorgione = Barbarelli' and 'Giorgione was so called because of his size' are both true, but replacement of the name 'Giorgione' by the name 'Barbarelli' turns the second statement into a falsehood: 'Barbarelli was so called because of his size'.14 Sitoilarly, if we hold as true the statements: 'the'Morning Star=Venus', 'the Evening Star= Venus', and 'the Morning Star=the Evening Star' it seems reasonable that we can substitute 'Venus' and 'the Morning Star' in some contexts. E&ut if we introduce second order predicates we get into trouble. Whereas it may be true to say: 'John believes that the Morning Star is Venus', it may not necessarily be true to say 'John believes the Evening Star is Venus'. Now, what Thieme has in mind here is a particular type of identification or bandhu 'tie; nexus' well known from the Brahmanas and Upanisads. Examples are: samvatsaroyajhah, 'the sacrifice is the year' (SatBr 11.2.7.1); adityo 'gnih, 'the [sacrificial] fire is the sun' (SatBr 10.5.4.1); atmaivedam sarvam, 'this entire universe is nothing but the Self (ChUp 7.25.1); akaso brahma, 'brahman is Space' (ChUp 3.18.1). The task of the Brahmanas, is to explain ritual and its implements, including formulas, acts, and material substances, by discovering their real nature and establishing their internal relations. To account for these items, their purpose and meaning in ritual, these identifica14

Example from Qmne 1969:139.

Substitution

181

tions were made. Later Vedic literature found itself facing a fundamental question: if the Veda is a body of injunctions, how can it be meaningful? First and foremost this question pertained to the problem of meaning in ritual. How can the Upanisadic statement that 'all this is brahman' be true when it is against a lot of Vedic injunctions? The first aim of the Upanisads was accordingly to give these ritual injunctions a meaning, and in doing so they employed the familiar type of identifications within a metaphysical frame. It seems clear to me that in thi? context the 'is' of these identifications lends itself to the meaning 'stands for; represents' rather than '='. It may well be the case that it was these very same identifications which laid the foundation for the model of substitution as met with for example in the technical literature of grammar. It is clear that they are referred to by the term ddesa (hence Thieme's claim that ddesa means 'Substitution' as well as 'Substitut'?), but, as far as I can see, there is no textual evidence which warrants the conclusion that ddesa means 'substitute' or 'substitution' rather than 'specification; teaching; instruction' or 'Anweisung; Vorschrift' in the Brahmanas and oldest Upanisads. Indeed, some of the examples adduced by Thieme seem rather to speak against his own case. The phrase athdta ddesah: neti neti (BAUp [M] 2.3.11 [K] 2.3.6) Thieme (ibid.:264) interprets as follows: 'Nunmehr seine (des makrokosmischen und mikrokosmischen Seelenwesens: purusa) Ersetzung (die seine mystische Wahrheit enthullt). Sie lautet immer wieder: "nicht'V15 This seems to me rather forced, and ddesa would make perfect sense as 'specification; teaching': 'now hence the teaching: neti neti". In other words, I think Thieme's conclusion is wrong. This is supported also by the following observation expressed by Wezler (1972:7): Nachdem P. Thieme und - unabhangig von ihm - Y. Ikari16 gezeigt haben, daB ddesa- m. auch in den Brahmanas und alteren Upanischaden die dem Paniniya aus der Astadhyayi gelaufige Bedeutung 'Substitution, Substitut' hat, wtirde man erwarten, bei Uberpriifung der Belege in den Kalpasutras zu dem gleichen Ergebnis zu kommen, zumal in diesen Werken im Zusammenhang detaillierter Angaben iiber den Vollzug bestimmter Opferhandlungen zahlreiche, und zwar konkrete 'Ersetzungen' gelehrt werden. Diese Erwartung wird freilich enttauscht. Tatsachlich namlich scheint adesa- in den Kalpasutras, wie bereits L. Renou festgestellt hat,17 nur im Sinne von a) 'Anzeige, Angabe' und von b) Anweisung, Vorschrift' vorzukommen.18
15

'Now his (of the macrocosmic and microcosmic soul-being: purusa) replacement (which dis16 closes his mystical truth). It is again and again: "not".' Y. ikari 1969. 17 Renou 1941/42:130. 18 'After P. Thieme and - independently of him - Y. Ikari have shown that ddesa- m. has the meaning "substitution, substitute", which is known to the Paninlya from the Astadhyayi, also in the Brahmanas and the older Upanisads, one would expect to reach the same result if one were to check the references in the Kalpasutras because m these works concrete "replacements" are taught in connection with detailed accounts of the execution of particular sacrificial acts. This expectation is not met. In fact ddesa- seems to appear in the Kalpasutras only in the sense of (a) "declaration, information" and (b) "order, rule".'

182

Indian semantic analysis

It would be highly curious if ddesa were to mean 'substitute'"or 'substitution' in the Brahmanas and oldest Upanisads, but not have such a meaning in the ritual Sutras which do operate with the notion of substitution in a technical sense. The surprising fact that we do not meet with this meaning in the Kalpasutras once again warrants the conclusion that Thieme was wrong or at least that he was reading too much into his textual evidence. In fact, I believe that it is perfectly possible to go directly fromlthe meaning and usage of the term ddesa in ritual texts to understand its use in the sense of 'substitute; substitution'. In ritual works ddesa means 'instruction; specification'. It is a specific injunction, a special application, which overrules what is already there, what is 'in place', sthdne. In other words, something automatically applies (prdpnoti) unless there is some specific instruction, ddesa, to overrule it. In practice this comes down to 'substitute', and the usage of the term ddesa in grammar is accordingly nothing more than a special application of its liturgical use. Let the following passage from the Baudhdyana Srautasutra suffice to buttress this claim (BSS 24.8): athdta ddesakdritdni vydkhydsydmo '19 'nddisto 'gmr ' api tu yathaitad bhavaty adhidevane juhoti rathamukhe juhoti... gdrhapatye juhotity ' anddista ahavanlya eva hotavyam ' anddisto 'dhvaryur ' api tu yathaitad bhavaty unnetd juhoti pratiprasthdtd juhotity ' anddiste 'dhvaryunaiva hotavyam ' anddistam pdtram' api tu yathaitad bhavaty ahjalind juhoti surpena juhoti krsndjinaputena . . . ayasmayena kamandalund juhotity ' anddiste srucaiva hotavyam ' anddista samid ' ddesdd eva samidham jdniydd' vijhdyate ndsamitke juhuydd yad asamitke juhuydd yathdjihve 'nnam dadydt tddrk tat tasmdt samidvaty eva hotavyam ' anddista upasamddhdyaiva hotavyam '// 8// Next, therefore, we shall explain the operations of specific instruction (ddesa). The [Ahavanlya] fire has not been specified. But it is the case that it is [specifically stated] as follows [in Brahmana]: 'he sacrifices [ghee] on the gambling ground';20 'he sacrifices on [the fire above] the front of the chariot';... 'he sacrifices on the Garhapatya fire'. [Yes, but] when there is no specific instruction {anddiste) [as to what it is], the sacrifice must be made on the Ahavanlya fire only. The Adhvaryu21 has not been specified. But it is the case that it is [stated] as follows: 'the Unnetr22 sacrifices'; 'the Pratiprasthatr23 sacrifices'. [Yes, but] when there is no specific instruction (anddiste), the sacrifice must be made by the Adhvaryu priest only. The sacrificial vessel has not been specified. But it is the case that it is [stated]
19

20

21 22 23

Caland, sensibly, has retained the sandhi of the manuscripts and only separated the sentences by placing a little stroke above the line, cautiously adding (BSS, vol. l:xm) 'without being sure, however, to have given always the right interpunction'. I retain his punctuation here. SatBr 5.4.4.22. The subsequent instances should, at least in principle, be similarly traceable quotations. I take the form of the passage to be a dialogue, a feature characteristic of early Sastnc prose. The Yajurveda priest (rtvij) in charge of all the manual actions of the sacrifice such as cooking the oblations, offering the oblations into the fire, etc. * One of the assistants of the Adhvaryu, in charge of extracting and pouring the Soma. The primary assistant of the Adhvaryu. '

Substitution

183

as follows: 'he sacrifices by means of the cupped hands'; 'he sacrifices by means of the Surpa (wickerwork basket for winnowing grain)'; 'he sacrifices by means of the vessel of black antelope skin' . . . 'he sacrifices by means of the bronze watervessel'. [Yes, but] when there is no specific instruction (anddiste), the sacrifice must be made by means of a Sruc24 ladle only. The kindling has not been specified. One should understand the kindling from the specific instruction (adesa) [to perform a sacrifice] itself. One should understand that one cannot make a sacrifice on a fire which has not already received kindling. To sacrifice on [a fire] without kindling is just like giving food to a person without a tongue; therefore one must sacrifice only on [a fire] which has kindling. When there is no specific instruction (anddiste), the sacrifice must be made only after the kindling has been provided. To sum up, unless there is a specific instruction, the default case is that one sacrifices on the Ahavanlya fire; unless there is a specific instruction, the sacrifice must be made by the Adhvaryu priest; unless there is a specific instruction, the sacrifice must be made by means of a Sruc ladle; unless there is a specific instruction, the sacrifice must be made only after the fire has received kindling. Otherwise a specific instruction (adesa) would tell you to substitute that for the general instruction. The Baudhayana Srautasutra is generally considered the oldest among the Srautasutras.25 The section from which the above passage stems, the karmdnta section, is a set of paribhdsa-like instructions, that is to say, a section that contains general rules of interpretation applicable to the science of ritual. The very methodology employed in this section is a methodology which requires the substitutional model. In fact, that the usage of adesa in the sense of 'substitute' is no more than a special application of its liturgical use in the sense of 'specific instruction', is thus brought out by the general methodology employed in the technical literatures of ritual and grammar, to which I shall now turn. The method of description in the literatures of ritual and grammar The notion of something automatically applying unless there is a specific instruction to overrule it reflects the general methodological division in Sastra between the general (samdnya) and the special (visesa). This division is crucial to the method of description met with in the literatures of grammar and ritual from Baudhayana onwards. To get a clear picture of how this methodology works, let me first of all turn to a familiar relation, namely, the relation between
24 25

The generic name for the three sacrificial ladles, the juhii, the upabhrt, and the dhruvd On this assumption W Caland (1903:11), following the general concensus of his time that Apastamba belonged to the fifth or fourth century BCE, gave the sixth century BCE as terminus post quern for Baudhayana. However, the dates of the Srautasutras are to be treated with a lot of caution, as noted by J. Gonda (1977.476-7): 'Whereas some authors wisely confine themselves to statements such as "the general period of the sutras extends from the sixth or seventh century before Christ to about the second century", others, overlooking the necessity to study the complex of this literature as a whole, could not resist the temptation to settle the age of a definite work within narrow limits.'

184

Indian semantic analysis

map and territory. If we have a map - and I think it is justified to call the ritual and linguistic descriptions of the ancient Indians a map - it will tell us a great deal about those who made the map. Reminding ourselves that the map is not the territory, we may then ask what features of the territory are represented on the map. If the territory is absolutely uniform, nothing would be represented on the map except the borders of the territory. Otherwise, what will be represented on a map is really differences of various kinds - differences in height, vegetation, surface, population structures, etc. Let the following passages from the ritual Sutras serve to illustrate how the basic framework of this methodology works: ASS 1.1.2—3: darsapurnamdsau tu purvarn vydkhydsydmas tantrasya tatrdmndtatvdt, 'but we shall explain the new and full moon sacrifices first since the basic model itantrd) has been handed down in that context'. ASS 2.1.1: paurnamdsenestipasusomd upadistdh, '[the rules for] the isti26 animal, and Soma sacrifices have been taught by [the rules for] the full moon sacrifice [which has been described already]'. AGS 1.7.1-2: atha khalilccdvacd janapadadharmd grdmadharmds ca tan vivdhe pratiydtl yat tu samdnam tad vaksydmah, 'now, the rules of different regions and towns are indeed varied, [and] those one must observe at the wedding; what is common [to them], however, we shall [now] describe'. SSS 1.16.1: vydkhydtau darsapurnamdsau prakrtir istipasubandhdndm, 'the new and full moon sacrifices which have been explained are the prototype for the isti and animal sacrifices'. BhSS 5.17.1-3: punarddheyam vydkhydsydmah I tasydgnyddheyena kalpo vydkhydtahl vikdrdn anukramisydmah, 'we shall [now] explain the renewal of the sacrificial fires; its ritual procedure has been explained by the [ritual of] setting up the sacrificial fires; we shall outline [only] the differences'. BhSS 6.15.4—5: dmdvdsyam tantram bhavatil tatraiso 'tyantapradesahI sarvesv istipasubandhesu ddrsapaurnamdsikd dharmd anuyanti, 'the new moon sacrifice is the basic model; in this respect the general rule is this: in all istis and animal sacrifices, the rules follow those prescribed for the new and full moon sacrifices'. BhSS 14.1.1-3: madhyamdinam savanam vydkhydsydmah I tasya pratahsavanena kalpo vydkhydtahl vikdrdn anukramisydmah, 'we shall [now] explain the midday pressing [of Soma]; its ritual procedure has been explained by the morning pressing; we shall outline [only] the differences'. Thus, the methodology employed revolves around such concepts as prakrti 'prototype' and vikrti 'modification'. This is also referred to in terms of an image from the art of weaving as tantra 'warp' and dvdpa 'wqof denoting respectively the basic model which is the constant part of a ritual and the
26

An isti is a type of Srauta sacrifice which requires vegetable offerings in contrast to the bloody animal sacrifices.

Substitution

185

special features which differ from one ritual to another. The pattern is clear: the general situation or the shared features are described first, and the modifications or changes or differences are later introduced as specific instructions, adesas. Indeed, this is a procedure which is commonly resorted to in any science. Turning to Panini and his grammar, we shall first of all have to ask a fundamental question: what language did he make a map of? The answer to this question is well known, but absolutely essential. At one and the same time he deals with all linguistic facts known to him as Sanskrit. These linguistic facts were, on the basis of stylistics, divided into two main areas: chandas, the language met with in the corpus of Vedic texts, and bhdsa, the standard colloquial Sanskrit of his time and clearly a spoken language. The rules of the AstddhydyT apply simultaneously to both these domains in as much as the rules applicable only to the Vedic domain are stated by explicitly mentioning the Vedic domain as a particular domain within the total domain, the totality of linguistic facts.27 Moreover, differences within this total domain are accounted for in various ways, notably, regional, social, and pragmatic usages - for instance, there are rules which specify which forms to use when contempt is to be expressed. Panini also refers to easterners and northerners, probably to account for dialectal usages. By way of example, rule A 6.3.32 mdtarapitardv udicdm teaches that according to the northerners mdtarapitarau is a valid form (as against mdtdpitarau), and A 8.2.83 pratyabhivdde 'sudre teaches that in returning a greeting, but not to a Sudra, the last vowel of a sentence becomes pluta and gets the acute accent. Variation is also accounted for by rules stating options. It was thus quite natural for Panini in his description of language to adopt the same methodology that we meet with in the ritual Sutras. One starts out with what is common to all domains of language and gives a description by means of rules which apply to it entirely. Differences of various kinds are then accounted for by rules giving exceptions or options. Accordingly, Panini formulated general rules (utsarga) to account for features shared by all domains, and exceptional rules (apavdda) to account for deviating features pertaining only to a particular domain. This is what gives us patterns such as: generally item X occurs, but in chandas item Y (or: item X or Y); generally form X occurs, but to express contempt one may use form Y; preferably form X occurs, marginally form Y may be used instead; generally the suffix X is added but
27

As noted already (p. 32 above), Yaska seems to treat chandas and bhdsa m much the same manner (Nir 2.2): athapi bhdsikebhyo dhdtubhyo naigamdh krto bhdsyantel damundh ksetrasddhd ml athapi naigamebhyo bhdsikdhl usnaml ghrtam iti, 'moreover, Vedic prima nouns are analysed on the basis of verbal roots belonging to classical Sanskrit, such as damundh, ksetrasddhdh, but also classical [primary nouns] on the basis of Vedic [roots], such as usnam, ghrtam\ But although chandas and bhdsa were not conceived of as different stages of language ordered in time, it is still clear that they were considered different linguistic domains. So, for example, the fifth-century grammarian Bhartrhan in his MahdbhdsyadTpikd (MbhD 288,17-18). asya tu chdndasatvdd bhdsdydmprayogo duhsampddah, 'but because th is a chandas form, it is difficult to use it in bhdsd\

186

Indian semantic analysis

'

under certain circumstances suffix Y occurs instead. The picture is that of a linguistic universe where different elements in given contexts or at given points in the process of derivation may become manifest or actualised at some location in grammatical space. And the idea of where something is and where something could be also leads naturally to the notion of substitution. Returning to the image of map and territory, I may quote M, Deshpande (1985:124; emphasis in the original): Theoretically speaking, Panini's conception of Sanskrit is that of a panchronistic flatland. It is neither purely diachronic nor purely synchronic. It includes all known diachronic and synchronic facts of Sanskrit. From the point of view of comparative linguistics this might be considered an apt description, but it is nevertheless imposing one important notion onto Panini's model, namely the notion of time. Panini's description of Sanskrit comprises all linguistic facts known to him but nowhere does he.say anything like 'earlier one said X, but today one says Y' or anything that cpuld be interpreted in that direction. In other words, his flatland is not panchronic to him. Panini's model has nothing to do with chronos at all. This also means that nothing like language development or linguistic change through time is accounted for. The language he describes is timeless, later to be .termed nitya 'eternal' or siddha 'established'. In the tradition of Indian grammatical thought we meet with a fundamental debate on the nature of grammar and hence on the nature of words or speech elements. One view holds that they are eternally established, constant {nitya), while according to the alternative view they are produced (kdrya), as if processed by grammar. The discussion comes up at several places in the Mahdbhdsya. It is first met with in the Paspasdhnika in the context of Pataiijali's remarks on Rgveda 4.58.3 (Mbh 1:3,15-16): catvdn srngd trayo 'sya pddd dve sirse sapta hastdso asya I tndhd baddho vrsabho roravlti maho devo martydm a vivesa II Four horns, three feet are his, two heads, seven hands are his. Bound in three ways the bull roars repeatedly: a great god has entered mortals. As one might expect, the verse in question has been subject to widely differing interpretations which need not concern us here. The hymn itself praises the sacrificial ghee, which is first identified with Soma which has the form of a bull. According to Patarijali, the great god is sabda, Speech, the four horns are the four classes of words (nouns, verbs, preverbs, and particles), the three feet are the three tenses, etc. What is of immediate interest here, however, is that the two heads are seen as the two aspects of the nature of words or speech units: eternal (nitya) and produced (kdrya). It is the nitya view which gets the attention in Vyakarana, the kdrya view is to my knowledge never really developed at all. When it comes to the model of substitution, the Sanskrit grammarians who

Substitution

187

believed in the doctrine of nityatva were forced to adopt a model of substitution where substitutes are considered substitutes of whole words. This is first stated in a verse quoted in the Mahdbhdsya (Mbh 1:75,13-14): sarve sarvapadddesd ddksiputrasya pdninehl ekadesavikdre hi mtyatvam nopapadyate II All [the substitutes taught] by Panini,28 the son of Daksl, are substitutes of whole words. For eternality {nityatva) is incongruent with change in a part. Of course, the Sanskrit language, like all languages, developed and changed through time. The interesting question is whether the Indian grammarians were aware of such changes and, if they were, how they dealt with the phenomenon. Indeed, how is it possible to account for what we would call historical change within a non-historical model? It is clear that the grammarians after Panini were aware of differences and variations within different domains of Sanskrit usage, but they never adopted any idea of historical change. The question has been elegantly answered by Deshpande (1985), and I shall not go into details here. Briefly, the idea is that the Vedas are eternal and infinite, and that we consequently do not know everything that is in them. This makes it possible to claim that new forms or new linguistic usages do not exist. Patanjali does indeed encourage the inclusion of previously unattested forms. Since the language is eternal, such usage can only be rediscovered old usage. This principle was adopted also in other Sastras, such as Dharmasastra. A good example is the practice of burning a sati or virtuous widow, a practice which is nowhere prescribed or even mentioned in authoritative literature. But since the practice exists, it must have been dealt with in a lost or inaccessible part of the Veda. With respect to grammar, the situation has been described by Deshpande (1985:137): From the theoretical point of view of this eternal Sanskrit, no new facts of language can come to light because there can be no new facts. Only previously unknown facts can become newly known. Thus, the theory of eternality of Sanskrit separates the existence of language from its attestation in history. The question is then who has the power to determine a newly discovered fact as correct usage. This is the cultural elite known as the Sistas, the model speakers of Sanskrit. In the words of Deshpande (1985:143—4): Each new generation of these linguistic elites may provide previously unknown facts about this eternal language. Thus the grammar of this eternal language is in theory, quite paradoxically, not eternal. It has to be a continually changing entity ... It is a living organism which, in theory as well as in practice to some extent, continues to grow with the language and ever remains aware of and inclusive of newly discovered facts about this language.
28

The genitive form pdnineh could also indicate that it is Panini's point of view that is stated here* 'In the opinion of Panini, the son of DaksT, all [substitutes] are substitutes of whole words.'

188

Indian semantic analysis

The Indian grammarians' model of language, then, is a map of where something is and where something could be. The task of grammar isto structure the linguistic material and suppress the occurrence of incorrect forms. Regional or other usages, which were accepted as correct, would not appear as more than acceptable substitutes which occur in the place of other forms under given circumstances.29 At the level of grammatical operations the substitution procedure implies that in the process of a given derivation, a linguistic item X is replaced by another item Y in some specific context. Such an operation accounts for the fact that X does not occur in its possible domain of applicability, Y occurring there instead. One more feature of the substitutional model requires mention. In the AstddhydyT, rule A 1.1.56 sthdnivad ddeso 'nalvidhau lays down the important principle, known as the sthdnivadbhdva principle, that the substitute is treated like the substituend except in a rule teaching an operation that depends on sounds.30 The same principle applies in the ritual Sutras, as stated in the Bhdradvdja Srautasutra (BhSS 6.15.7): tad yatra prdkrtasya yajndngasya sthdne vaikrtam dmanet tat tasya dharmdn labhetl yathd saramayam barhir barhisah sydmdkd vnhTndm sthdlT kapdlasya caruh puroddsasyeti II When someone teaches a modification to occur in the place of a sacrificial requisite pertaining to the prototype, it takes on the characteristics of the latter. For example, grass of the sara type that of barhis, sydmdka grains that of paddy, a vessel that of potsherds, and boiled rice that of a cake. Similarly, the Apastamba Srautasutra clarifies the situation arising when some sacrificial substance is lacking in full or in part (ApSS 24.3.52-4): sistdbhdve sdmdnydt pratimdhihl taddharmd ca sydtl mdtrdpacdre tacchesena samdpnuydtl
29

30

P Kiparsky (1979) argues that the three different words used by Panim to indicate optionality, vd, vibhdsd and anyatarasydm, did m fact serve to distinguish between preferred usage, marginally acceptable usage and equally acceptable options respectively The^ later grammarians treat them all as synonyms indicating mere optionality If Kiparsky is right, this requires an explanation If the grammarians have to account for the total, timeless domain of eternal Sanskrit, it is clear that they could no longer easily view options as graded possibilities. Grammar has to account for all correct usage pertaining to this eternal language A form is thus either valid or not - it cannot be more or less valid It seems to me very unlikely that Pamnlyas after Pamni did not notice, for example, that vd 'or' could be interpreted differently from vibhdsd which Pamni defines ( A l l 44) as 'na vd' On the one hand, if the na here is taken to negate vd, one would simply have to accept that something negated is different from itself unless it is zero, as for example a feature of height in a complete flatland. On the other hand, it is perfectly possible that it means 'or not' rather than 'not or', in which case matters would no longer be so straightforward However, the possibility remains that the Pamnlyas after Pamni let the idea of nityatva weigh so heavily that they preferred to disregard any distinction between the three terms If, again, Kiparsky is right, this would also indicate that the doctrine of nityatva was not firmly rooted in the grammatical climate of Pamni The rule and Patanjah's remarks on it have been studied in detail by S D Joshi and J A F Roodbergen 1985 and 1990

Substitution

189

When the prescribed [sacrificial requisite] is lacking, a substitute (pratmidhi) [must be used] on the basis of similarity. And [this substitute] would possess the characteristics of that [which it replaces]. If there is partial lack [of the sacrificial requisite], one should finish [the ritual] by means of what is left of it. So the sthdnivadbhdva principle applies also in the ritual domain, either to the substitute which has become necessary because of the method of description, that is to say, when the prototype specifies some sacrificial implement which is to be replaced by another in the modification sacrifice, or it applies because a certain substance prescribed for a certain ritual is not available and another substance has to be used in its place. Ritual implements can be replaced by other implements, and again we are not talking about identity between them but about semantic equivalence. The sthdnasasthi: Panini's rule A 1.1.49 and its indigenous interpretations It is the genitive case ending which marks the specific relation of 'being in the place of. As mentioned already, the grammarians refer to this genitive as the sthdnasasthi. By way of example, rule A 2.4.52 aster bhuh [drdhadhdtuke A 2.4.35] teaches that bhu replaces as in the domain of drdhadhdtuka suffixes.31 This could be written: asteh sthdne bhuh, 'bhu [occurs] in the place of as [in the domain of drdhadhdtuka suffixes]'. We have seen that the expression tasya sthdne 'm the place of it' is normal Sanskrit usage, and the use of the genitive to indicate a substituend in the AstddhydyTis accordingly nothing more than employing established linguistic usage for a technical purpose. In the Astddhydyi the, sthdnasasthi is defined by rule A 1.1.49 sasthi sthdneyogd, and an inquiry into the notion of grammatical I substitution and the use of the genitive case to indicate it would thus reason\ ably begin with an investigation of this rule and its indigenous interpretations. But in the same way as the investigation of the relation sustained between a term and its corresponding explanatory expression in the Nirukta had to proceed through nirvacanas formulated in a highly technical manner since they alone could be expected to yield sufficiently systematic information, an investigation into the sthdnasasthi and the relation sustained between the terms involved in a substitution operation has to begin by investigating the rule which teaches the technicalities of substitution. Accordingly, only an investigation of A 1.1.49 and the literature pertaining to it can in the first instance be trusted to ^provide us with reliable constants in the investigation of the sthdnasasthi and the model of substitution. An investigation of Panini's notion of substitution through internal analysis
31

A 3 4 113 tinsit sdrvadhdtukam assigns the name sdrvadhdtuka to the finite verbal suffixes (tiN) and those marked with an anubandha, diacritic marker, S A 3 4 114 drdhadhdtukam sesah then assigns the name drdhadhdtuka to the rest, that is, to all other suffixes added to a verbal root

190

Indian semantic analysis

of the AstddhydyT alone is hardly likely to be very fruitful, and it is not only Panini's model of substitution ^hich is of interest here. As iriiportant for the present study is the way the subsequent Paninlya tradition conceived of it. Accordingly, I shall take as my basic text the Mahdbhdsya discussion of A 1.1.49, together with its commentaries. This provides the mostjsubstantial discussion not only of the rule itself but also of more wide-ranging issues pertaining to the notion of substitution in general. Patanjali's remarks on A 1.1.49 have been summarised briefly by Ojihara and Renou (1962), although these authors are primarily concerned with the Kdsikdvrtti of Jayaditya and Vamana and with its commentaries, Haradatta's PadamanjarT and Jinendrabuddhi's Nydsa. P.-S. Filliozat (1976) has done a French translation of the Mahdbhdsya on A 1.1.49, together with Kaiyata's PradTpa and its subcommentary, the Uddyota, by Nagesabliatta. Filliozat's work is part of a large project aiming at a complete translation of Patanjali's work with the two commentaries, and as such is not too much concerned with details pertaining to the issues I wish to discuss. The present study requires a detailed study of Patanjali's remarks on A 1.1.49 together with a thorough discussion of issues raised in the commentaries and other relevant Paninlya literature. Only by considering in detail the issues raised therein, the arguments advanced and the implications they entail, can we hope to get the full picture of the Paninlya model of substitution, and to find out whether Paninlyas differ among themselves and, eventually, why they prefer one possible solution to another. Only by such a procedure would it be possible to understand the full import of the simple phrase 'X in the place of Y' in the Paninlya context, to < determine what this 'place' is and the relationship between the three terms, involved in a substitution operation. And only by thoroughly investigating s these questions would it be possible to find out whether the notion of grammatical substitution is compatible with the semantics of nirvacan'a analysis as laid > down by Yaska and employed in the discourse of subsequent Sanskrit litera-; ture. Survey of Patanjali's remarks on rule A 1.1.49 The series of arguments advanced by Patafijali in his remarks on rule 1.1.49, frequently involve other rules and examples which require explanation.] Moreover, in what follows I shall consider several issues at length in the light; of what the commentators have to say. To some extent this makes the presenta-^ tion wind back upon itself, and a brief survey of Patanjali's discussion may! { therefore be useful. * \ As noted already, a genitive case suffix is defined in rule A 1.1.49 by the expression sthdneyogd. Patanjali's first task is to question this expression, the| mere form of which creates problems. He justifies its interpretation as a com-^ pound, more precisely as a bahuvnhi compound whose members are not in| syntactic agreement. The irregular occurrence of the locative case suffix,

Substitution

191

which may even be considered short for the instrumental case suffix, is accounted for on the grounds that we are dealing with a ready-made form, a nipdtana. Under this interpretation rule 1.1.49 states that a genitive case suffix marks a relation (yoga) which is in some way defined by the sthdna, a term to which the tradition has attributed different senses. The next issue in Patanjali's discussion is the purpose served in formulating the rule. In his first vdrttika Katyayana suggests that the rule is stated in order to restrict the number of possible interpretations of a genitive case suffix. According to Patanjali, a hundred meanings may be expressed by means of a genitive case suffix, and all of these would apply when a genitive case suffix is pronounced. Hence, a restriction is to be made that in the science of grammar > one should consider a genitive case suffix to be a sthdnasasthi and nothing else. The second vdrttika rejects this idea, arguing that if every genitive were to be interpreted as a sthdnasasthi, then other usages of the genitive, such as the avayavasasthT, the genitive which marks a part, would no longer be [ accounted for. As examples are adduced the genitive forms sdsah and gohah of A 6.4.34 sdsa id anhaloh and A 6.4.89 ud upadhdyd gohah. Only when ; these forms are interpreted as partitive genitives do the rules mentioned take \ effect properly. The third vdrttika tries to meet the objection raised by the second in claimling that the restriction would apply only when there is doubt about the relation ^involved and that there is no doubt in the case of genitives such as the partijtive. Patanjali declares that even this need not be stated, and illustrates the ^situation with an example from everyday life. When someone asks the way [from one place to another, only that information is given to him that would remove doubt. Similarly, in grammar, a restriction applies only when there is idoubt. At this point in his discussion Patanjali mentions two other interpretations of the expression sthdneyogd, namely sthdne 'yoga and sthdne yogavatl Sender the first alternative a genitive case suffix would be identified as a sthdnasasthi when no particular relation is specified by context. Under the i second alternative a genitive case suffix would be identified as a sthdnasasthi pWhen there is a possibility that it denotes a variety of relations. ? Leaving these proposals aside with no further remarks, Patanjali then turns : to the fourth vdrttika which suggests that it is rather a specially distinguished Igenitive case suffix which is defined as sthdneyogd. That is, the problem of •Identifying a sthdnasasthias such could be solved by adding to it some particular mark by which it would be easily recognised. Such a procedure would |iowever create other problems. Patanjali quotes A 6.4.34 sdsa id anhaloh and | a 6.4.35 sd hau to illustrate the difficulties entailed by such a marking device. Jlp rule 6.4.34 the item sdsah is understood to exhibit a partitive genitive. Now, the same item sdsah is considered to recur in the subsequent rule 6.4.35 sd hau Jwhere it no longer exhibits a partitive genitive. Instead the required results can pe obtained only if the genitive is interpreted as a sthdnasasthi. But in rule "6,4.35 the item sdsah is merely understood, it is not mentioned. Its only

192

Indian semantic analysis

mention is in rule 6.4.34 where it exhibits a partitive genitive. So, Patanjali objects, the recurrence of the item sdsah in rule 6.4.35 gives us a mere understanding of the word form sdsah, and to mere understanding a mark cannot be added. To get around this difficulty he suggests that the mark be added to the ddesa, the substitute, instead. Challenging the whole claim that rule 1.1.49 states a restriction, Patanjali at this point shifts the argument to cases where one single genitive case suffix serves to mark several different relations. This, of course, creates problems. As a possible way out Patanjali proposes that if no restriction is made at all, a genitive case suffix may mark various relations according to context. In that case, however, an earlier problem arises at one remove, namely that there are a hundred meanings expressed by a genitive case ending, and all of these would apply when a genitive is pronounced. This, says Patanjali, i!s no problem. We are dealing here not with relations in general, only with those that could possibly obtain between one linguistic item and another. But even then, he has to admit, such relations as adjacency, dnantarya, and nearness, sdnupya, may obtain in addition to the one denoted by the sthdnasasthT. Doubt will thus arise anyway. Condemning this as 'mere doubt', Patanjali invokes a paribhdsd, general principle of interpretation, which states that in all cases of doubt, one particular conclusion should be ascertained by reasoned explanation. By force of this paribhdsd Patanjali concludes that when interpreting a genitive case we should supply the locative word form sthdne. But if this is the core of the matter, then why should rule 1.1.49 be stated? It needs to be stated, Patanjali claims, in order to ensure that whatever ends in a genitive case suffix, specifically any form after which the suffix is actually pronounced, should be linked with the sthdna. Interpreted in this way, says Patanjali, rule 1.1.49 serves to provide another paribhdsd according to which substitutes occur in the place of what has been expressly enunciated. Various conclusions may be drawn from Patanjali's remarks on rule 1.1.49, but no purpose would be served in considering these before tatanjali's discussion has been examined in more detail. The expression 'sthaneyoga'" Patanjali begins his discussion of rule 1.1.49 with an explanation of the expression sthdneyogd, the mere form of which raises certain difficulties. Judging from the feminine suffix -a it would have to be a compound, qualifying sasthl But the locative case ending of the element sthdne would generally preclude an interpretation as a compound. Patanjali proposes t^ie following analysis of the expression (Mbh 1:118,6-7): kirn idatn sthdneyogeti I sthdne yogo 'sydh seyam sthdneyogd I saptamyalopo nipatandtll trtiydyd vaitvaml sthdnena yogo 'sydh seyam sthdneyogd II What is this 'sthdneyogd' ? [As a bahuvnhi compound qualifying sasthi it can be analysed as:] that of which there is a relation (yoga-) in the sthdna (sthdne), that is

Substitution

193

sthdneyogd. Since this is a legitimate form (nipdtana), there is no elision of the locative case suffix. Or there is e instead of the instrumental case suffix (-end), that of which there is a relation with the sthdna (sthdnena) is sthdneyogd. Patanjali interprets sthdneyogd as a bahuvnhi compound which qualifies sasthT, and accounts for the non-elision of the locative case suffix on the grounds that we are dealing with a legitimate form, a nipdtana, a specially listed, ready-made form.32 A 2.2.24 anekam anyapaddrthe accounts for the bahuvnhi compounding of more than one word in order to denote something different, that is, different from what is directly referred to by any of the individual words which appear in the compound. In the present instance the compound qualifies the sasthT, the sixth case suffix. Now, a compound is considered a nominal stem according to A 1.2.46 krttaddhitasamdsds ca which teaches that also a derived linguistic unit which ends in a krt or a taddhita suffix or is a compound is called a nominal stem, prdtipadika.33 According to A 2.4.71 supo dhdtuprdtipadikayoh, which teaches elision by the element luk34 of a case suffix that occurs in a root or in a nominal stem, the locative case ending should be elided in the compound sthdneyogd. There are, however, exceptions to this rule. A 6.3.1 alug uttarapade states that in the forms listed in the subsequent rules there is no elision by luk of a case suffix before the last member of a compound. A 6.3.9 haladantdt saptamydh samjndydm, for example, teaches that a locative case ending is not elided before the last member of a compound when the prior member is a stem ending in a consonant or in a, provided the whole compound is a name. Rules A 6,3.10-20 teach similar specific conditions for the construction of an aluksamdsa where, in particular words, the locative case suffix is retained. There does not seem to be any specific rule to prevent elision of the locative case suffix in a compound such as sthdneyogd. So Patanjali relies on Panini's authority and states that there is non-elision of the locative case suffix 'since this is a legitimate form', nipdtandt. As for the alternative interpretation, sthdnena yogo 'sydh seyam sthdneyogd, which introduces the instrumental case suffix instead of the locative with the statement that the item -e has replaced -ena, the same nipdtana argument would hold.35 The reason for introducing an alternative analysis which involves
32 33

See Renou 1955. The previous rule, A 1 2 45 arthavad adhdtur apratyayah prdtipadikam, states that a nomina stem, prdtipadika, is a meaningful linguistic unit which is neither a verbal root, nor a suffix, nor a form which ends m a suffix However, we are not really dealing with a proper nominal stem here because of the feminine 34 By A 11 60 adarsanam lopah, elision (lopa, lit 'loss') is defined as 'non-seemg' A 11 61 pratyayasya lukslulupah then lists luk, slu and lup as three specific names of elision which apply only to the non-seemg of suffixes. 35 Or rather, there is elision of the third case suffix, but the -a of the stem is replaced by -e as it is before the dative and ablative plural ending -bhyah according to A 7,3 103 bahuvacane jhaly et which teaches that e replaces the final a of a stem before a plural ending beginning with a consonant. Clearly this rule is not applicable here, and hence, it is argued, we are dealing with a nipdtana

194

Indian semantic analysis

the instrumental case may quite simply be that the word yoga is equated with sambandha 'relation', and that these terms are normally construed with the instrumental case so that the meaning of the compound sthdneyogd could be expressed as: 'that of which there is a relation with the sthdna\ On this view the sthdna itself is one of the two terms of the relation. But thp instrumental case could also imply that a relation obtains which is caused by tfiie sthdna. The locative, on the other hand, could express that the relation is 'in the sthdna\ in the sense that one term is related to another term with the sthdna as a necessary condition for the relation. Thus the alternatives have quite different implications for the interpretation of the relation involved. * It seems to me likely that the locative word form sthdne is retained on the basis of this form's use in ordinary language, while the instrumental would more properly account for the needs of the present context. As we have seen, the expression tasya sthdne 'in the place of it', where the locative form sthdne is construed with a genitive form, is normal Sanskrit usage. P.-S. Filhozat (1976:364) translates sthdneyogd as 'qui exprime une relation par 1'occurrence'36 regardless of whether it is the locative or the instrumental case which is involved, claiming that 'dans les deux cas le sens de compose est le meme, la troisieme et la septieme desinence pouvant avoir le sens de cause'.37 It is certainly true that in Sanskrit both the instrumental and the locative can express the sense of cause, but one may ask why Patanjali should bother to introduce an alternative explanation if it did not add any new dimension at all. By the expression 'par 1'occurrence' Filliozat has interpreted the word sthdna as prasanga 'possible appearance; chance' or 'occurrence'. According to the grammatical tradition this is a fully legitimate interpretation, although it is not the only one possible. A detailed examination of the complexities pertaining to the interpretation of the relation in question and the traditionally accepted meanings of the word sthdna will be carried but below. Concerning the interpretation of the compound sthdneyogd itself, a look at the commentaries is illuminating. Kaiyata remarks as follows (MbhP I:408a): kim idam itil samdse vibhaktisravandprasangahI asamdse 'pi 'yuktih' itiprdpnoh, na tu yogeti prasnahll 'What is this [sthdneyogaP* The issue is this, if [sthdneyogd is] acompound, then the hearing of a case suffix should not take place38 But if it is not a compound, [the form] 'yuktih'' would occur [in the feminine] and not 'yogd\ I take this to mean that the form yoga is acceptable when it appears as the final member of a compound, but when there is^ao^compound, yuktih would be the acceptable feminine form. The stem yoga- is formed by addition of the
36 37 38

'Which expresses a relation by occurrence ' In both instances the meaning of the compound is the same, in as much as*both the third and the seventh case endings can have the sense of cause' That is, there should be elision of the locative case suffix according to A 2 4 71 supo dhdtupratipadikayoh

Substitution

195

suffix GHaN {-a with initial accent and guna or vrddhi). This suffix is added to a verbal root to express bhdva 'being; mere activity' in the form of an action noun according to A 3.3.18 bhdve or to express a kdraka other than the agent in lexicahsed derivatives39 according to A 3.3.19 akartan ca kdrake samjhdydm. The corresponding feminine would be formed according to A 3.3.94 striydm ktin which introduces the feminine suffix KtiN (-ti with initial accent and weak grade) for any root in the senses taught by rules 3.3.18 and 3.3.19. The resulting stem would thus be yukti,40 Nagesa accordingly remarks on Kaiyata's statement (MbhU I:408a): yuktir in I sasthiti stnlingavisesanatvdd iti bhdvah, '"yuktih [would apply]": since it qualifies the feminine gender of [the word] sasthT; that is the idea'. And Bhartrhari, in the same context, remarks as follows m his DTpikd (MbhD 288,10-12): kim idam sthaneyogeft / samdsdsamdsayor yugapadasambhavdt saptamisravandd asamdso 'yarn41 sydtl yoge 'pi42 tdpa updddndt samdsah sydtl itarathd hi yuktir iti prdpnotity aha kim idam sthaneyogefr / 'What is this sthdneyogd"?' Because it is impossible for something to be a compound and a non-compound at the same time, this would be a non-compound since we hear the locative case ending. But from the use of the feminine suffix TdP m the word yoga-, [that is, in yoga,] it would be a compound. For otherwise the word yuktih would apply [as the correct feminine form]. So he says: 'what is this sthdneyogdr Later in the same passage (MbhD 288,15 f.) Bhartrhari comments on Patanjah's alternative analysis by means of the instrumental case. He seems to refer, through the word dcche, to Panmi's rule A 7.1.39 supdm sulukpurvasavarndccheyddddydydjdlah which accounts for various case endings in Vedic (chandas), thinking the -e (se) mentioned in this rule to be the one which has replaced -ena. Provided I have understood his reference correctly, this seems rather awkward. Bhartrhari himself realises the difficulty of such a formation since the rule is taught for Vedic language.43 Nagesa too remarks on the nipdtana issue and the instrumental case alternative (MbhU 1:408), but I do not consider it worthwhile to enter into details here. In brief, he states that in a compound with a word ending m the instrumental case there is e for a when a third triplet ending, that is, an instrumental
39

That is, words whose meanings are not transparent from the application of grammar alone but ascertained through the lexicon of the Sanskrit language On the sense of samjfid, see G B Palsule 1966 40 The derivation of a feminine form from rules A 3 3 18 and A 3 3 19 is blocked by A 3 3 94, see Vaidyanatha's Chdyd (Bhargavasastrl JosT 1951 408, note 4) tathd cdkartanti prdptaghafiah striydm ktind bddha iti bhdvah, 'and thus, by the rule striydm ktin, there is blockin of GHaN which would apply according to the rule akartan etc , that is the idea' 41 So read with Palsule and Bhagavat (1991 18) and MbhDms fol 200b for MbhD asamdso sydt 42 MbhDms fol 200b clearly reads pi but adds what seems to be a superfluous va, Palsule and Bhagavat (ibid) read iti ca 43 MbhD 288,17-18 asya tu chdndasatvdd bhdsdydm prayogo duhsampddah, 'but because th is a chandas form, it is difficult to use it in bhdsd'

196

Indian semantic analysis

case ending, is elided by luk. This is recognised as a nipdtana rather than taught as a substitution 'in the place of the third' because the -ina substitute would occur first, because there is no authority for blocking it as an antaranga operation,44 because there is no authority for blocking the elision by Ink, and also because vrddhi would occur.45 Nagesa makes it clear that if the substitution in question were of the case suffix, then Patanjali should have said i-tvam instead of e-tvam. So Nagesa tells us how he thinks this nipdtana works and why Patanjali calls it e-tvam rather than i-tvam, and explains why the e itself is the nipdtana. But just like the other commentators, Nagesa does not give us a clue as to why an interpretation of the compound involving the instrumental case is brought up at all. Some commentators give only the trttydvigraha, the analysis by means of the instrumental case; thus, notably, two prominent commentators on the Siddhdntakaumudi, Bhattoji Diksita46 in his autocommentary Praudhamanoramd,41 and Jnanendrasarasvatl48 in his TattvabodhinT.49 The evidence available makes me stick to my previous suggestion: an analysis by the instrumental is quite simply prompted by the fact that one would expect the word yoga, which is considered equivalent to sambandha 'relation' by the commentators, to be used with an instrumental sthdnena in the present context.50 i Commentators thus follow Patanjali in regarding the expression sthdneyogd as a compound. The retention of the locative case suffix precludes an interpretation of it as a samdnddhikaranabahuvnhi, a bahuvrihi compound whose members are in syntactic agreement. If it were such a compound, rule 1.1.49 would teach that a sixth triplet ending, that is, a genitive case ending, is used when the relation is the sthdna. Rather, the compound has to be interpreted as a vyadhikaranabahuvnhi, a bahuvrihi whose members are not in syntactic agreement.51 The basic meaning of rule 1.1.49 would thus be: 'a genitive case ending is used to mark a relation which is in some way defined bf the sthdna". In the course of his remarks oh rule 1.1.49, Patanjali mentions two other interpretations of the expression sthdneyogd. Since these to a certain extent rest X 44 That is, the operation which gives e for a +1. This operation is considered to be antaranga, an operation whose conditions or causes of application are internal or proximate with regard to those of another operation which accordingly is termed bahiranga, whose conditions are external. In the process of formation an antaranga rule is present in the mind prior to a bahiranga rule, and is thus considered stronger than it. See p. 215 below. 45 By rule A 6.1.88 vrddhir eci, -a and e in sthdna e would be replaced by ai. 46 Bhattoji probably flourished in the last half of the sixteenth century and the early seventeenth 47 century; see Cardona 1976:284. Ed. V.L. Joshi 1966:70L^ 48 Jnanendrasarasvatl was a contemporary of Bhattoji according to Yudhisthira Mlmamsaka (1973:412). Other scholars have assigned him to the early eighteenth century; see Cardona 1976, note 475. 49 SKI:44. 50 The assumption is supported by Patanjali's final remarks on A 1.1.49 (Mbh 1:119,26): sasthyantam sthdnena yathd yujyeta0; see p. 218 below. 51 This is the view also of Nagesa (MbhU I:408b), further, by way of example, the Kdsikdvrtti (Kas 1:170); the Bdlamanoramd and TattvabodhinT commentaries on the Siddhdntakaumu (SK 1:44); Praudhamanoramd (ed. V.L: Joshi 1966:70).

Substitution

197

upon arguments brought up during his discussion, they will be dealt with as they appear. The purpose of rule A 1.1.49 Having thus analysed sthdneyogd as a bahuvrlhi compound, Patanjali continues his discussion of rule 1.1.49 by questioning the purpose of the rule. This serves to introduce the first vdrttika (Mbh 1:118,8—13): kimartham punar idam ucyate I sasthydh sthdneyogavacanam niyamdrthamll 1 // niyamdrtho 'yam drambhahl ekasatam sasthyarthd ydvanto vd te sarve sasthydm uccdritdydm prdpnuvantil isyate ca vydkarane yd sasthisd sthdneyogaiva sydd iti tac cdntarena yatnam na sidhyatlti sasthydh sthdneyogavacanam niyamdrthaml evamartham idam ucyate I I Now, for what sake is this [rule] stated? The statement of a relation defined by the sthdna for the genitive has as its purpose the making of a restriction/ vt. 1 This undertaking is to make a restriction. One hundred52 are the meanings for the genitive case suffix53 - or as many as there are - and all of them would apply when a genitive case ending is uttered. But we want the genitive case ending in the science of grammar to be the one which marks a relation defined by the sthdna only, and that does not happen without effort. Therefore [the vdrttika is made]: 'the statement of a relation defined by the sthdna for the genitive has as its purpose the making of a restriction'. So for such a purpose it is stated. Thus the first vdrttika claims that the rule is stated in order to establish a niyama, a restriction with regard to the possible interpretations of a genitive case suffix in the rules of grammar. Kaiyata remarks on this in the following way (MbhP I:409a): ekasatam iti I sasthTdandakapdthdsrayenaitad uktam II ydvanto veti I sabde ydvatdm sambhava ity arthahll 'One hundred': this is stated by relying on the SasthTdandakapdtha. 'Or as many as there are': that is to say, as many of which there is a possibility in a word. Nagesa's commentary here is quite important for the interpretation of the compound sasthyarthdh (MbhU I:409a): sasthldandakapathah - granthavisesahll bhdsye - sasthyartha iti I sasthiprayojanakds tadarthdnvayinah sabddh satam ity arthahllyad vd sasthyarthdndm satatve tesdm satatvam arthasiddham bodhyamll ekasatam iti nirnayendbfiidhdya punah samdigdhdbhidhdnam ay uktam ata aha — sabda iti I Normally ekasatam means 'hundred and one', e.g. ekasatam adhvaryusakhah, 'hundred and one Adhvaryu (= Yajurveda) branches' (Mbh 1:9,21). Nagesa takes it in the sense of satam 'one hundred'; see below. 53 For a detailed discussion of the compound sasthyarthdh m the light of Nagesa's commentary, see below.
52

198

Indian semantic analysis

'SasthXdandakapathd' - [namely] a particular text.54 In the Bhdsya [Patanjah states] 'sasthyarthdh'. That is to say, those words whose purpose [for use] (prayojana) is [the expression of] the genitive case ending and which are invariably accompanied by its meaning, are a hundred. Or else, if the property of being a hundred pertains to the meanings [expressed] by the genitive case suffix, then one should recognise that the condition of being a hundred for those words is established through their meanings. Moreover, when it has been stated with decision 'one hundred', an ambiguous expression is inappropriate. Therefore he [=Kaiyata] says: 'in a word'. Thus Nagesa suggests two interpretations of the compound sasthyarthdh in accordance with two senses attributed to the term artha, namely 'purpose; motive (prayojana)9 and 'meaning'. Under the first alternative, sasthyarthdh is considered a bahuvnhi compound which qualifies sabddh 'words'. It seems reasonable to assume that this is a straightforward samdnddhikaranabahuvrlhi, a bahuvnhi compound whose members are in syntactic agreement. The sense would thus be that there are a hundred words whose 'purpose' (prayojana) is the genitive case ending. The words alluded to injhis way are svasvdmin 'possessed-possessor', avayavdvayavin 'part-whole', etc. These are further said to be tadarthdnvayin 'accompanied by the meaning of it', that is, by the general meaning of the genitive case suffix, namely 'relation'. To get the full idea of this it is necessary to investigate more closely the sense of prayojana in the present context. Since Nagesa's remarks have implications of considerable importance, it seems worthwhile to go into details here. In grammatical literature the word prayojana generally means 'purpose; motive; object'. It is quite commonly the term used to refer to the necessary purpose of a rule. Grammatically, the word is formed as a derivation from the verb praAyuj 'use; employ', from which the equally common term prayoga 'use; employment' is derived as well. \ Nydyasutra 1.1.24 defines the term prayojana in the following way: yam artham adhikrtya pravartate tat prayojanam, 'that goal (artha) with regard to which one is active, that is prayojana' (NS 45). Vatsyayana, in his commentary on Nydyasutra 1.1.1, interprets prayojana as karanasddhana, that is, as a noun expressing the instrument of the activity denoted by the verb to \yhich the word is related: yena prayuktah pravartate tat prayojanam, 'that because of which the intent one is active, that is prayojana' (NS 5). According to this analysis, the word is formed with the suffix LyuT (-ana55 with guna and presuffixal accent) in the sense of karana 'instrument' by A 3.3.117 karanadhikaranayos ca. An analysis as karanasddhana would make it easy to distinguish the basic meanings of prayoga and prayojana. While the formeFcould be rendered 'use; employment', the latter would lend itself to an interpretation a& 'that through
54

55

Such a work is unknown to me It seems reasonable to assume that it woufd have to be a list giving the different words or meanings which determine the relations marked by a genitive case suffix, words like svasvdmin 'possessed-possessor', etc The element -yu is replaced by -ana according to A 71.1 yuvor andkau.

Substitution

199

which there is an employment of something', and hence be rendered 'purpose; motive', etc. The encyclopaedic Sanskrit lexicon Vdcaspatyam, an orthodox pandita work, does indeed give karanasddhana as the basic interpretation of prayojana:56 prayujyate 'nena, '[something] is employed etc. by means of if. The lexicon continues by quoting the Nydyasutra definition which, as seen, uses the term artha to define prayojana. Now, with regard to the compound sasthiprayojanakdh which qualifies sabddh 'words', one should note that it is common practice in grammatical literature to say that a rule X is the prayojana of a rule Y when rule Y is introduced first and one asks why it exists. This is also normal Sanskrit usage. So, just as one can say that rule X is the prayojana of rule Y when rule Y is introduced first, in the same way one can say here that the genitive case ending is the prayojana of certain words which come up prior to the utterance of the genitive case ending. Employing the samdsdnta kaP (-ka), one may thus construct a bahuvnhi compound and say that rule Y is X-prayojanaka, or that certain words are sasthiprayojanaka. The prayojana is that a genitive case ending is enunciated. If we apply a karanasddhana analysis of the word prayojana, then Nagesa's interpretation of sasthyarthdh as a samdnddhikaranabahuvnhi would be: 'a hundred words have as their purpose [for use the expression of] the genitive case ending'. It should be mentioned, however, that Nagesa in a different context interprets the word prayojana as karmasddhana, that is, as a noun denoting the object of the activity expressed by the verbal root *4yuj with the preverb pra. The context is the discussion of the compound dharmaniyamah which occurs in the very first vdrttika in the Paspasdhnika of the Mahdbhdsya. Among the various interpretations Patanjali suggests there (Mbh 1:8,4-6) is dharmaprayojano niyamah on which Nagesa remarks (MbhU 1:65b): sa eva dharmah, tena prayujyate - dksipyate iti karmalyudantahprayojanasabdahl sa cdsddhunivrttirupo niyamah/ This very [niyoga 'injunction'] is Dharma. Through it something is used, that is, implicated, and the word prayojana is accordingly formed by adding the suffix LyuT {-ana) in the sense of karman (object). And this, m the form of suspending incorrect [speech], is the restriction. The expression tena prayujyate would at first sight indicate an interpretation as karanasddhana, but tena 'through if refers rather to the preceding word dharmah here. The compound dharmaniyamah, understood as dharmaprajojano niyamah, would then mean that the restriction is something which is implicated through injunction (dharma = niyoga), and dharma would thus be the cause of the restriction. Applying this to the compound sasthiprayojanakdh, one would also have to take into consideration that this could be a vyadhikaranabahuvnhi, a
56

Taranatha Tarkavacaspati 1873-84.

200

Indian semantic analysis

bahuvnhi whose members are not in syntactic agreement. According to such an interpretation and an analysis of the word prayojana as karmasddhana, one would get the meaning: 'there are a hundred words which have something accomplished through the genitive case suffix'. Under the first of the alternative interpretations suggested by Nagesa, then, it is possible to conclude that ekasatam sasthyarthah means that there are a hundred words which have the genitive case suffix as their purpose for use and which are invariably connected with its general meaning 'relation'. Any one of these words would bring about a genitive case ending. But their meanings too would bring about the genitive. So Nagesa suggests another interpretation of sasthyarthah, now taking artha in the sense of 'meaning'. The compound would then have to be a tatpurusa, but taking it as a straightforward sasthitatpurusa in the sense of 'the meanings of the genitive case' is precluded by Nagesa's interpretation of the compound sthdneyogd as a vyadhikaranabahuvnhi and also by ,what Nagesa states elsewhere (MbhU I:408b): sthdnasdmipyddayah sasthyarthasambandhanirupakdh, na tu sasthyarthah, 'sthdna, sdrnipya, etc., these condition the relation which is the meaning of the genitive case, but they are not the meanings of the genitive case'. Under the second alternative suggested by Nagesa it is thus reasonable to interpret sasthyarthah as 'meanings for the genitive case' in the sense that there are one hundred words whose meanings serve to bring about the genitive case suffix. One could also get this by reasoning from anvaya and vyatireka, that is, from the continued presence and absence of units.57 The anvaya mode of reasoning is an analytic procedure which, formally expressed, states that when A occurs, B occurs; and when A is absent, B is absent. Thus, if these meanings occur, a genitive case ending invariably occurs. Nagesa's point here is that just as sthdna is not the meaning of the genitive case suffix, one cannot talk of a hundred meanings of the genitive. Rather, those words or meanings are a hundred which determine the relation denoted when a genitive case suffix is uttered and for which the genitive accordingly is a purpose for use. On the second alternative, Nagesa remarks, thfe property of being a hundred is arthasiddha, effected through meaning, as opposed to the first alternative where it would be sabdasiddha, effected through the forms themselves. To sum up, a hundred words, or their meanings, may serve to bring about a genitive case suffix. The words thus alluded to are 'possessed-possessor', 'part-whole', etc. These words or meanings determine the relation marked by a genitive case suffix. The problem, then, is that if any of these woMs or meanings occur, then a genitive case ending is brought aboutT Conversely, if a linguistic item is uttered with a genitive case ending, then all of these words or meanings would possibly apply if nothing is undertaken to prevent this from happening. The first vdrttika accordingly suggests that the purpose of rule
57

On anvaya and vyatireka, see Cardona 1967-8; 1981; 1983:37-116; 1988:499-503.

Substitution

201

1.1.49 is to establish a restriction in this respect so that in the science of grammar a genitive case suffix marks only that relation which is in some way defined by the word sthdna. Having thus suggested the purpose of the rule, Patanjali introduces the second vdrttika which points out that such a restriction is too rigorous, since it would exclude genitives such as the partitive etc. (Mbh 1:118,13-17): asti prayojanam etatl kirn tarhitil avayavasasthyddisv atiprasangah sdso goha itill 211 avayavasasthyddayas tu na sidhyantil tatra ko dosahl sdsa id anhaloh [6.4.34] iti sdses cdntyasya sydd upadhdmdtrasya cal ud upadhdyd gohah [6.4.89] iti gohes cantyasya sydd upadhdmdtrasya ca//5S This is the purpose. What then? There would be over-application with regard to the genitives which mark a part etc., such as sdsah and gohah.' vt. 2 But then the genitives which mark a part etc. are not accounted for. What fault is there? [A 6.4.34] sdsa id anhaloh would then [teach a substitution] in the place of the final sound of [the root] sds [when it is a presuffixal base] as well as in the place of any penultimate sound. [A 6.4.89] ud upadhdyd gohah would [teach a substitution] in the place of the final sound of [the presuffixal base] goh as well as in the place of any penultimate sound. The point brought up here is that if the purpose of rule 1.1.49 were to state a complete restriction with regard to the interpretation of a genitive case suffix met with in the rules of grammar, then other usages of the genitive, such as the partitive, would not be properly accounted for. In order to show how such an interpretation of 1.1.49 would affect operations taught by certain rules, Patanjali quotes as examples A 6.4.34 sdsa id anhaloh and A 6.4.89 ud upadhdyd gohah. Rule 6.4.34 provides for the substitution of the vowel / for the upadhd, the penultimate sound,59 of the root ^Isds ' order; instruct' when the latter is an anga, a presuffixal base,60 followed by aN or by any suffix beginning with a
58 59

See the Bhasya discussion at A 6.4.1, vt. 2 (Mbh 111:178,19-21). Although the term 'sound' may be considered to imply an idea of sequence, I have chosen it to render Sanskrit varna, and hence 'penultimate sound' to render upadhd, wishing to avoid the complicated concept of the phoneme. (See in this respect e.g. R. Harris 1981:57 f.) From v a practical point of view one may conveniently use the term 'letter', as done, for example, by Kielhorn (1874), but this refers to writing rather than to speech. By the term 'sound' I refer to \4 the basic individual sound-units listed in the Sivasutras, the sets of sounds preceding the actual body of rules in the AstadhyayT, others which are denoted by them, and the sounds termed ayogavdha (see below, p. 223, note 114). f° The not quite accurate but still preferable rendering 'presuffixal base' for the technical term anga I have taken from Cardona (e.g. 1976:187, et passim). The term is defined by A 1.4.13 - yasmdtpratyayavidhis tadddipratyaye 'ngam as that which begins with that linguistic element P to which a suffix is added when the suffix follows. The term vidhi generally means 'rule' as well as 'that which is taught in a rule'. The commentators interpret pratyayavidhih in A 1.4.13 as 'addition of a suffix'. This sense is easily established by rules A 3.1.1, 3.1.2 and 2.3.29. J.W. * Benson (1990) has carried out a thorough investigation of the term anga and its usage, based primarily on Patafljali's remarks on the term.

202

Indian semantic analysis

consonant and marked with a diacritic K or N.61 The words upadhdydh khiti recur from A 6.4.24 aniditdm hala upadhdydh khiti which provides for the elision of a penultimate n in a verbal root which ends in a consonant and is followed by a suffix marked with a diacritic K or N provided the addition of n is not signalled by a diacritic marker /. Rule 6.4.89 ud upadhdyd gohah provides for the substitution of the vowel u for the penultimate sound of the presuffixal base goh (^guh 'hide'), provided a suffix follows which begins with a vowel, act The latter requirement is taught by A 6.4.77 aci snudhdtubhruvdm yvor iyahuvahau from which aci recurs. If rules 6.4.34 and 6.4.89 are to work properly, the words sdsah and gohah have to be interpreted so as to exhibit partitive genitives, contextually bound to upadhdydh which exhibits a genitive to be interpreted according to rule 1.1.49. Thus sdsa upadhdydh teaches a substitution for the penultimate of sds, that is, a, and goha upadhdydh teaches a substitution for the penultimate of goh, that is, o. If, however, the genitive exhibited in sdsah and gohah were to be interpreted according to rule 1.1.49, they would no longer be contextually bound to upadhdydh and this term would thus erroneously effect the substitution of any penultimate sound. Likewise, sdsah and gohah would effect the substitution of the final sounds of sds and goh according to A 1.1.52 alo 'ntyasya which teaches that substitution takes effect on the final sound of the form indicated as a substituend by the genitive case. The words sdsah and gohah of rules 6.4.34 and 6.4.89 would thus signal the replacement of s and h by i and u, while the word upadhdydh would signal the occurrence of these replacements in the place of the penultimate sound of any item which is followed by the suffixes specified by the respective rules. This interpretation of the Bhdsya passage in question is in agreement with the remarks of the commentators. Kaiyata states the following (MbhP I:409b): avayaveftV niyamena sambandhdntardndm nivartandd iti bhdvahlj upadhamatrasyefrV sdsind sambandhdbhdvdtpacddisambandhino 'pity arthahl tatra hy aniditdm ity ata upadhdyd iti vartatell [He brings up the question of] 'part': because the restriction cancels the other relations [marked by a genitive case suffix]; that is the point. 'Of any penultimate': there being no relation with sds, the sense is [that substitution would occur] even for [the penultimate which is] the relatum in pac etc. For there [in A 6.4.34] there is recurrence of upadhdydh 'of the penultimate' from [A 6.4.24] aniditdm etc. The niyama, restriction, would follow from rule 6.4.24 into the present rule which contains the genitive sdsah: Nagesa adds to this (MbhU I:409b): tasya sthdnasasthitvenopadhdvisesane sdmartydbhdvddmbhdvah II Because this [= sdsah] ends in a sthdnasasthi, there is no capability for it to be a qualifier of the penultimate; that is the idea.
61

For example, the past participle sis-td 'ordered; instructed' is formed by adding the suffix Kta (-td) to the presuffixal base sds.

Substitution

203

That is to say, sdsah, and likewise gohah, would no longer be contextually bound to the genitive upadhdydh. Due to the niyama expressed as the purpose of rule 1.1.49 by the first vdrttika, they would instead have to be interpreted according to this rule. The result of this would eventually be that the substitutions taught by rules 6.4.34 and 6.4.89 would be directed to the penultimate sound of any item, such as pac etc., to which the suffixes stipulated by these two rules are added. Clearly, this is not Panini's intention. Accordingly, the claim of the first vdrttika that the purpose of rule 1.1.49 is to make a restriction cannot imply that any genitive encountered in the rules of grammar is to be interpreted tas denoting one relation only. So the second vdrttika objects to such an interpretation by pointing out that the rules of grammar also contain genitives such as the partitive and that these have to be accounted for as well, a fact which is amply illustrated by the two examples sdsah and gohah quoted by Patanjali. The third vdrttika tries to meet the objection raised by the second in claiming that the restriction is valid only when there is doubt about the relation (Mbh 1:118,18-25): avayavasasthyddindm cdprdptir yogasydsamdigdhatvdt II3 // avayavasasthyddindm ca niyamasydprdptih I kim kdranam I yogasydsamdigdhatvdt I samdehe niyamo na cdvayavasasthyddisu samdehahl kim vaktavyam etatl na hi I katham anucyamdnam gamsyate621 laukiko 'yam drstdntah I tad yathd I loke kamcit kascit prcchati grdmdntaram gamisydmi panthdnam me bhavdn upadisatv itil sa tasmd dcaste I amusminn avakdse hastadaksino grahitavyo 'musminn avakdse hastavdma itil yas tatra tiryakpatho bhavati na tasmin samdeha iti krtvd ndsdv upadisyate I evam ihdpi samdehe niyamo na cdvayavasasthyddisu samdehahlI * 'But [the restriction] would not apply to the genitives which mark a part etc., because there is no doubt with regard to the relation [in these cases].' vt. 3. But the restriction would not apply to the genitives which mark a part etc. Why? Because there is no doubt with regard to the relation. A restriction applies when there is doubt, but there is no doubt with regard to the genitives which mark a part etc. Is this to be stated? Definitely not. How can it be understood when it is not stated? From daily life there is the following illustration. Just as, in the world, when someone asks somebody: 'I am going to another village, could you point out the way to me?' This other one tells him: 'at that point you turn right, at that point you turn left.'63 Making out that there is no doubt with regard to a path going transversely there, he does not point that one out. And so here as well a restriction applies where there is doubt, but there is no doubt with regard to the genitives which mark a part etc.

?? On the future form gamsyate, see B. van Nooten 1970. &* The expressions hastadaksinah and hastavdmah are considered bahuvrihi compounds by th commentators. Thus Kaiyata states (MbhP I.410a). hasto daksino yasya, 'that of which the hand is right'. Since a path does not have parts or hands (cf. the Chdyd, quoted Bhargavasastrl JosT 1951:410, note 1: mdrgdnavayavatvdi), this is slightly expanded by Nagesa (MbhU I:410a): daksino hasto yasya sannihita ity arthah, 'that is to say, that [path] to which the right hand is near'. Puzzled by the unusual order of the members of these compounds, Bhartrhari (MbhD 290,14-17) explains them as being basically tatpurusa compounds and refers to Panini's rule A 2.1.30 trtiyd tatkrtdrthena gunavacanena, for which see S.D. Joshi 1969:168 ff.

204

Indian semantic analysis

The argument of the second vdrttika, that the restriction which was postulated as the purpose of rule 1.1.49 by the first vdrttika would be too rigorous in that it would affect the interpretation even of such genitives as the partitive, is here challenged by the claim that a restriction works only where there is doubt, and that such genitives as the partitive are beyond doubt as to their interpretation. As Kaiyata remarks (MbhP I:409b): aniyamaprasange niyamah kriyate, na tuprdg evdvasthite niyamell When there is a possibility that there would not be a restriction, a restriction is made, but not when a restriction has indeed been established already. Or, in the words of Nagesa (MbhU I:409b): avayavasasthyddindm istdndm nivrttaye niyamdprdptir iti vdrtikdrthah, 'the point of the [third] vdrttika is that a restriction would not come about to cancel such genitives as the partitive which are desired'. Patanjali claims that even this need not be stated, and compares the situation to one where a man wanting to know his way to some other village is told only of those points on his path where doubt may arise. The idea is, as I understand it, that one makes a restriction only with regard to such places as where there is a fork-shaped division of the path, not with regard to totally oblique paths. There is no doubt that a crossing path is not the path one is instructed to follow, and so nothing is said about it. Such a path is, according to Nagesa (MbhU I:410a), purvapascimdnyataradisi gacchato daksinottaramdrga iti ydvat, 'as much as a road going south-north for one who moves in the direction of either east or west'.64 Similarly, in grammar a restriction applies only when there is doubt, but there is no doubt with regard to genitives such as the one that marks a part, since these are established beyond doubt already. At this point in his discussion of rule 1.1.49 Patanjali mentions two alternative interpretations of the expression sthdneyogd. According to the first oii these it is to be read as sthdne 'yoga (Mbh 1:119,1): atha vd sthdne 'yoga sthdneyogd/ kim idam ayogetil avyakt^yogdyogdll Since the word yoga so far ha$ been taken in the sense of 'relation', the following translation seems reasonable:
64

In their edition of Bhartrhari's Dipikd Abhyankar and Limaye read as follows (MbhD 290,13-14): yau hi panthdnav bhavatas tatra na jndyate kasmin pqthibhdge grdmafy samnivista iti. In the Dipikd manuscript, however, the first part of this reads (MbhDms foL 202b): yau hi yamthdndm ekadikkau0. It is understandable that Abhyankar and Limaye take °dikkau as a bahuvrThi compound and, lacking a reference for this, emend the impossible yamthdndm to panthdnav0. Their emendation anekadikkau is possible, giving the sense of 'two paths going in a plurality of directions (not in two)' or, suiting my picture of the situation, 'two paths going in [slightly] different directions', however I fail to see why it is neces,n sary. I propose to read ekadikkau with the manuscript, a reading retained also by Palsule and Bhagavat (1991:20,26). Accepting the other emendations suggested by; Abhyankar and Limaye, one could then translate the passage as follows: 'For, when you hav6 two paths going in one direction, it is not known along which division of paths the town is located.'

Substitution

205

Or else sthaneyogd is [to be analysed as] sthane 'yoga. What is this ayogd! That [genitive case ending] which denotes a relation that is not clear. The remarks of Kaiyata seem to support the interpretation of yoga as 'relation' (MbhP I:410a): ayogeri/ yogam antarena sasthyd evdbhdvdd visisto yogo yasyd ndstiti sdmarthydt pratvyatel ayogd': since there is no genitive case ending without some relation, we must understand that [what is marked by the genitive defined by rule 1.1.49] is something for which there is no specified relation [established already]. Under this alternative sthaneyogd is no longer considered a compound. Instead it is taken as a phrase which is to be divided sthane 'yoga, where ayogd, in the sense of avyaktayogd 'which identifies a relation that is not clear', is considered a nansamdsa65 qualifying sasthT, while the word sthdna, in the locative, defines the relation marked by the genitive case suffix. This solution logically presents itself on the basis of the foregoing discussion in the Bhdsya. If the purpose of A 1.1.49 is to express some sort of restriction with regard to a genitive case ending met with in the rules of grammar, and if it is argued that this restriction must not affect genitives such as the partitive because these are established already and thus beyond doubt, it becomes possible to say, with Kaiyata, that no genitive case suffix occurs unless some relation is to be denoted, and that A 1.1.49 supplies the interpretation for a genitive form which, unlike sdsah and gohdh discussed above, is not contextually bound. In other words, where no particular relation is specified, the restrictive rule 1.1.49 applies to mark the relation in question as defined by the locative word form sthane. This solution is the only one presented in the Bhdsdvrtti of Purusottamadeva,66 a simple commentary on the AstddhydyT, where, incidentally, the 'word ayogd is glossed avyaktasambandhd,61 the word yoga thus specifically identified as synonymous with sambandha 'relation'. , In Nagesa's Uddyota we meet with a considerably more sophisticated interpretation of the sthane 'yoga alternative and of the word yoga in particular. This seems worthy of mention regardless of whether it represents tenets valid for Patanjali. Nagesa makes the following remarks on Kaiyata's interpretation of the issue (MbhU I:410a): visista itil nirnayavisista ity arthahl ayam eva bhdsye vyaktasabdasydrthah I atra pakse yogasabdah karmavyutpattyd sambandhiparahl 'visistaK: that is to say, specified through ascertainment. This alone is the sense of vyakta in the Bhdsya. On this view, by its analysis as object, the word yoga means relatum. $ That is, a compound with the negative particle as its first member. In the sense of avyaktayogd, the expression ayogd is taken as a compound of the sakapdrthiva type by Sivaramendrasarasvatlin his Ratnaprakdsa commentary on the PradTpa (ed. M.S. Narasimhacharya 1975:302). | 6 Generally placed in the early twelfth century; see Cardona 1976:282. ? Ed. Dwarikadas Shastri 1971:9.
1

206

Indian semantic analysis

This means that when the word yoga is analysed as karmasddhana, it denotes 'something which is related' and not 'relation'. Thus, when sthdneyogd is interpreted as sthdne 'yoga, rule 1.1.49 teaches, according to Nagesa, that a genitive case ending which, in contrast with genitives such as the partitive, has an unspecified relatum marks a relation defined by the word sthdne. Patanjali goes on to present still another possible interpretation of the expression sthdneyogd (Mbh 1:119,2-3): atha vd yogavatT yoga I kd punar yogavatT/ yasya bahavo yogdhl kuta etatl bhumni hi matub bhavatill Or rather yoga is [to be interpreted as] yogavatT. What, then, is [the sense of] yogavatil That for which relations are many. How does this come about? Because the suffix matUP applies in the sense of abundance. Under this alternative sthdneyogd is to be read as two separate words, sthdne and yoga, the latter formed by adding the secondary suffix aC {-a) to the stem yoga- according to A 5.2.127 arsaddibhyo }c which teaches that the suffix aC is added to the words arsas etc. in the sense of matUP, and by then adding the feminine -a (TdP). According to A 5.2.94 tad asydsty asminn iti matup, the possessive suffix matUP (unaccented -mat, under certain conditions to be replaced by -vat) is added to form derivatives in the sense of 'that is pertaining to it' or 'that is located in it'. A vdrttika quoted by Patanjali under this rule states that the suffix can only be added when certain meanings, such as censure or abundance, are to be expressed.68 The gana, list of words, starting with arsas is an dkrtigana, an open-ended gana. Under this alternative, then, yoga means 'abundant in relations' and rule 1.1.49 would thus teach that a genitive case suffix which may denote many relations marks the relation defined by the word sthdne. I am not fully confident of the text at this point, but Bhartrhari seems to take the following stand in his Dipikd (MbhD 291,1-3): atha va yogavati yogeft / sarvasydh sasthyd yogo69 'stiti10 bhummti vyapadisyate111 yaivdnekdrthasambandhini saiva yoga12 yogavatil yd tu nirjhdtayogd sd tu ndyogd ndpf3 yogavatiti sthdne na bhavisyatil 'Or rather yoga is [to be interpreted as] yogavati' All genitives mark [some] relation, [and] therefore it is pointed out that [matUP applies] 'in the sense of abundance'. But only that [genitive] which marks a relation with multiple meanings is yoga, that is, yogavatT. But [a genitive] which marks a relation that is fully known,
68 69 70 71 72 73

See the Bhdsya discussion under the first vdrttika on A 5.2.94 and the verse cited there (Mbh I 393,15-16) With MbhDms (fol 202b), against the emendation yoge of Abhyankar and Limaye (MbhD) MS (ibid.) reads stiff. Emended, also by Palsule and Bhagavat (1991 21,12), from MS (ibid ) reading \yadisyate. MS (ibid.) reads yoge. So read with Palsule and Bhagavat (1991.21,13), MS (ibid.) reads sd tu ndyogd ndmapi0, Abhyankar and Limaye (MbhD 291,2-3) lead* °sd tu nal yogdndm api°.

Substitution

207

[that genitive] is neither ayogd 'marking a relation which is not clear' nor yogavatT 'abundant in relations', and hence does not apply to mark a relation defined by the sthdna. Nagesa, who sticks to his interpretation of yoga as 'relatum', offers an important remark here which however bids some difficulty as to its interpretation (MbhU I:410a): yogavatlft' pakse 'py evam eva, sthdnarupasambandhiniyamdrthatvddasya// 'yogavatf: even on this view things are just the same, since [the rule] has as its purpose a restriction on the relatum in the form of the sthdna. Now, this is ambiguous. The compound sthdnarupasambandhiniyama0 can either be understood so that sthdnarupa qualifies the word saynbandhin, the sthdna thus becoming a relatum of the relation in question, or it can be understood so that sthdnarupa qualifies sambandhiniyama, the restriction thus appearing in the form of the sthdna. It makes sense to refer to the sthdna as a relatum, directly, less so to refer to it as a niyama, directly. The first alternative therefore seems to me the most plausible one, although the second cannot be excluded completely. Under the first alternative the statement would mean that when there is a possibility of many relata, rule 1.1.49 serves to restrict a genitive to mark a relation with a specific relatum only, namely the sthdna. Under the second alternative there is a restriction on the relatum, the restriction being the sthdna itself. Taking into consideration that the word sthdna appears in the locative in the phrase sthdne yoga, this could be taken to imply that rule 1.1.49 comes into play to mark a relation which has one relatum or, possibly, both relata restricted to appear in the sthdna. The two possible interpretations may thus be seen to represent two different views of how the relation marked by a sthdnasasthT is to be considered. On the first view we would be dealing with a relation where there is a restriction that one of the relata has to be the sthdna. On the second view the sthdna would not be a relatum but something which restricts the one relatum or both relata of the relation in question. A full discussion of the problems connected with the exact determination of the relation marked by a sthdnasasthT will be carried 4 out below. Suffice it here to say that these two views come up in that context, as does the problem that the sthdna, strictly speaking, is not a relation. This may be the reason for Nagesa's choosing to interpret yoga as 'relatum' and not as 'relation' when sthdneyogd is not considered a compound. To sum up, when the wording sthdneyogd is split, Paninlyas leave us with two options. The phrase can be read either as sthdne 'yoga or as sthdne yoga. Generally speaking, both of these interpretations specify certain conditions for interpreting a genitive as a sthdnasasthT. According to the first, when there is an uncertain relation or, following Nagesa, an uncertain relatum, the relation marked by the genitive is defined by the word sthdne. According to the second, when there is an abundance of relations or, following Nagesa, an abundance of

208

Indian semantic analysis

relata, one is similarly to supply the word sthdne. None of these alternatives is considered by Katyayana, and Patanjali seems to present them only in passing, leaving them as mere possibilities arising through the claim that the purpose of the rule is to make a restriction. Patanjali now turns to the statement of the fourth vdrttika which brings up an entirely new solution to the problem raised by the second vdrttika of how to identify a sthdnasasthi when also genitives such as the partitive are to be taken into consideration as possible interpretations when a genitive case suffix is met with (Mbh 1:119,4-15): visistd vd sasthi sthdneyogd II All atha vd kimcil litigant dsajya vaksydmTtthamlingd sasthi sthdneyogd bhavatltil na ca tal lingam avayavasasthyddisu karisyatell yady evam sdsa id ahhaloh sd hau sdsigrahanam kartavyam sthdneyogdrtham lingam dsanksydmiti I na kartavyaml yad evddah purastdd avayavasasthyartham prakrtam etad uttdratrdnuvrttam sat sthdneyogdrtham bhavisyatil kathaml adhikdro ndma triprakdrahl kascid ekadesasthah sarvam sdstram abhijvalayati yathd pradipah suprajvalitah sarvam vesmdbhijvalayati I aparo 'dhikdro yathd rajjvdyasd vd baddham kdstham anukrsyate tadvad anukrsyate cakdrenal aparo 'dhikdrah pratiyogam tasydnirdesdrtha iti yoge yoga upatisthatel tad yadaisa pakso 'dhikdrah pratiyogam tasydnirdesdrtha iti tadd hi yad evddah purastdd avayavasasthyartham etad uttaratrdnuvrttam sat sthdneyogdrtham bhavisyati I sampratyayamdtram etad bhavati I na hy anuccdrya sabdam lingam sakyam dsanktuml evam tarhy ddese tal lingam karisyate tat prakrtim dskantsyati II Or rather, a distinguished genitive case suffix [signifies] sthdneyogd.' vt. 4 Or else, having added some mark, I shall state that a genitive case suffix with such a mark is established as sthdneyogd. And this mark will not be made for such genitives as the partitive. 4 > If so, mention must be made of; sdsl, [the root ^Isds, the genitive^form of which recurs] in [A 6.4.35] sd hau [from A 6.4.34] sdsa id anhaloh, so I may add a mark to signify the sthdneyogd. [Such a mention] need not be made. That very [mention of sdsl] which is put forward in the context of the earlierjrule [A 6.4.34] to signify a partitive genitive, that [mention of sdsl] is recurrent in the subsequent rule [A 6.4.35] for the purpose of signifying the sthdneyogd. How? Becausean adhikdra74 is of three kinds. A certain kind, standing at its one place, illumines the whole corpus of rules as a bright-shining lamp fully illumines a house. Another [kind of] adhikdra is dragged along by means of the particle ca 'and', just like a log is dragged along which has been tied to a rope or an iron chain. Another [kind of] adhikdra makes its presence in rule after rule so that it need not be stated in each subsequent rule. Therefore, when the view is that a governing rule need not be stated in each subsequent rule, then, in fact, that very [mention of a genitive form], which in an earlier rule signifies a partitive genitive, is recurrent in a subsequent rule where it appears to signify the sthdneyogd.
74 4

Strictly speaking, an adhikdra is a governing rule, but Patanjali seems to usp the term in the wider sense of any statement that can be read into a rule and thus govern its interpretation; see below.

Substitution

209

[But] this brings about mere knowledge. For it is not possible to add a mark if the linguistic item is not pronounced. If it is like that, then the mark will be made on the substitute, and it will then leap over to the substituend75 [which is indicated by the genitive form recurrent from the earlier rule]. The fourth vdrttika suggests that the problem of identifying a genitive as a sthdnasasthlcould be solved by adding to those genitives some particular mark by which they would easily be recognised. But such a procedure creates certain difficulties. Patanjali quotes two rules to make this clear. The first is rule 6.4.34 sdsa id anhaloh, discussed earlier. In brief, this rule signals the replacement by i for the penultimate sound, upadhd, of the root ^Isds 'order; instruct' when the latter is a presuffixal base, anga, followed by aN or by any suffix beginning with a consonant and marked with a diacritic K or N. Now, the genitive form sdsah is understood to exhibit a partitive genitive, since the word is contextually bound to upadhdydh which recurs from A 6.4.24 aniditdm hala upadhdydh kniti where upadhdydh exhibits a sthdnasasthl. The second rule quoted by Patanjali is A 6.4.35 sd hau which teaches that the presuffixal base sds is replaced by sd before the imperative ending hi to give sd-dhi. The genitive form sdsah of rule 6.4.34 is understood to recur in rule 6.4.35 where it no longer signifies a partitive genitive contextually bound to upadhdydh, but a sthdnasasthl. In rule 6.4.35, then, the genitive sdsah is considered to recur. But it is not mentioned. Its only mention is in rule 6.4.34, but there it signifies a partitive genitive. In order to argue against the necessity of repeating sdsah in rule 6.4.35, ^Patanjali refers to three ways in which an adhikdra works. Strictly speaking, an adhikdra rule is a heading rule which governs subsequent rules in a particular section of the grammar. The adhikdra is thus an item which recurs {anuyartate) in each rule belonging to this section. But parts of rules are also considered to recur in subsequent rules, and these recurring items evidently fall fcnder the scope of an adhikdra as well, if one is to judge from Pataiijali's disgussion here. Moreover, the first kind of adhikdra mentioned by Patanjali in |his passage actually refers to that type of rule which is technically termed paribhdsd, general principle of interpretation, not restricted in its application |o the domain of a certain section but valid throughout the grammar. Rule 11.49, for example, is considered a paribhdsd rule.76 The second kind of

$S The exact sense of the term prakrti is a point of ambiguity here. Generally it means 'norm; original' or, in grammar, 'base; stem'. Panini does not employ prakrti as a technical term in such a sense, only prakrtyd which denotes a form that has not yet undergone any change (yrtti). In the present context prakrti could refer to the base sds, or to an earlier mention, prdkrta, that is, to what has come down from an earlier rule which in this case would imply that prakrti refers to the form sdsah as it has come down from rule 6.4.34, or it could refer to the sthdnin, the substituend. Kaiyata takes it in the latter sense; see below. $ Kaiyata remarks (Mbhp 1:41 la): pdrdrthyasdmydt paribhdsd ypy adhikdra ucyate, 'even paribhdsd is called an adhikdra because it is equivalent in serving the purpose of other [rules]'. In his commentary here Nagesa objects to the claim that a paribhdsd illuminates the whole

210

Indian semantic analysis

adhikdra mentioned here works by means of the particle ca 'and; also' which signals the recurrence of the item in question. The first and second kind of adhikdra may be excluded in the present context simply because they do not apply. It is the third kind of adhikdra which is relevant here. This is said to work in such a way that it 'makes its presence in rule after rule so that it need not be stated in each subsequent rule'. Both major governing rules and parts of rules which are to recur in subsequent rules were probably marked with a svarita vowel to be heard in the recitation of the AstddhydyT. This feature was lost early in the tradition.77 This marking device is accounted for by A 1.3.11 svaritenddhikdrah which states that an item characterised by a svarita vowel is to be considered an adhikdra. Patanjali's way of telling how the third type of adhikdra works is an echo of the first vdrttika on this very rule: adhikdrah pratiyogam tasydnirdesdrthah, 'the point oft an adhikdra is that it need not be stated in subsequent rules'. As Cardona (1968:448, note 4) has pointed out; the term adhikdra is here best interpreted as karmasddhana, that is, as denoting the item which is made to recur.78 Thus, in the present context, the item sdsah which is considered to recur in rule 6.4.35 from rule 6.4.34 seems to be regarded as an adhikdra of this third kind. Accordingly, it is only necessary to mention the genitive form sdsah once, so that by its mention in rule 6.4.34, although it there signifies a.partH tive genitive, it makes its presence in rule 6.4.35 where it signifies af sthdnasasthi.19 To this the objection is raised that the recurrence of sdsah in rule 6.4.3§ gives a mere understanding of the word form sdsah from rule 6.4.34, and to, mere understanding no mark can be added. How could one add a mark to aj

Footnote 76 (cont.) grammar, pointing out that there are two views concerning how a paribhdsd works (MbhU| 1:411 a). He discusses the issue at length m the context of panbhdsds 2 and 3 in hij Panbhdsendusekhara (Kielhorn 1871-4:6-26; ed. Abhyankar 1962:1-6). Concerning thS difference between a paribhdsd and an adhikdra, the Bhdsya discussion at A 2.1.1 samarthaM padavidhih betrays a certain confusion in this respect and tcrsome extent an overlapping between the two concepts. Discussing whether this rule is a panbhasa or an adhikdrM Patanjali here postulates a difference between the two in claiming that a paribhdsd is of tl|| first kind referred to above, whereas an adhikdra is stated to make its presence in rule aftei rule so that it need not be stated again in each subsequent rule, and thus equal to the third kiag referred to above. See Mbh 1:359,4-7 with Kaiyata's commentary. See also Mbh 1:70,20-11 where the concept of paribhdsd seems to include that of adhikdra. 77 Cardona (1968) has discussed Panini's use of svarita; references to other discussions perl taming to this topic are given by Cardona 1976:192. 78 The Kdsikd at A 1.3.11 glosses adhikdra by the word viniyoga (roughly, 'assignment; conn mitment'), an interpretation as bhdvasddhana, that is, as a noun expressing the mere activip of assigning, according to the PadamanjarT (KasP 1:408). Commenting on the Bhdsya a A 1.3.11, Kaiyata remarks that the word adhikdra may be interpreted either as bhdvasddhana§i l as karmasddhana (MbhP II: 145a). 79 Nagesa (MbhU 1:41 la) claims that we are dealing here with a sabdddhikdra, an adhikafM where one item of a rule goes into a subsequent rule in the sense that only the form goes dowi|p not the meaning. It should however be kept in mind that the major adhikdpa rule A 6.411 angasya exhibits a genitive to be interpreted by context, in rule 6.4.34 as a partitive geiutiv|f| in rule 6.4.35 as a sthdnasasthi. The genitive angasya is coreferential with sdsah in both rul$i

Substitution

211

word form that is not pronounced? Kaiyata articulates the problem clearly (MbhPI:411a): svaritatvapratijndndd uttaresu yogesu tasya sabdasydnumdnam bhavatu, lingdsangas tv anuccdritasya katham sydd ity arthahll That is to say, by the convention of recognising the existence of svarita vowels, this item [sdsah] should be inferred in subsequent rules80; but how is a [phonetic] mark added to something which is not pronounced? In order to meet this difficulty, Patanjali suggests the mark be added to the ^substitute, adesa. Through the relation that obtains between substitute and sub"stituend, sthdnin, the sense conveyed by the mark will be transferred to the subIstituend, marking it thereby as exhibiting a sthdnasasthl Kaiyata explains this in the following manner (MbhP I:411a): ddesalingdsangdt tasya ca sasthyabhdvdt tatsambaddhasthdnino lingakdryam phalisyatity arthahll That is to say, by adding the mark to the substitute, and by the fact that this does not exhibit a genitive case suffix, the operation [signalled] by the mark will take effect on the substituend which is related to the substitute. Thus, if a mark were added to the substitute sd in A 6.4.35 sd hau, this mark gvould direct the operation to take effect on the related substituend expressed jjby the recurring genitive form sdsah which is thus identified as exhibiting a hthdnasasthiIn the context of A 6.4.35. The mark would not direct the operaEon to the substitute sd itself, since this item does not exhibit a genitive case Sending. I Challenging the whole claim that the purpose of rule 1.1.49 is to state a flpstriction, Patanjali next points out the difficulties of cases where there is only p e genitive case suffix but several things to be qualified through it (Mbh Sfil9,16-19): yadi niyamah kriyate yatraikd sasthy anekam ca visesyam tatra na sidhyatil angasya halah anah samprasdranasyeti I hal api visesyo 'n api visesyah samprasdranam api visesyam I asatipunar niyame kdmacdra ekayd sasthyanekam l visesayitum I tad yathdl devadattasya putrah pdnih kambala itil tasmdn ndrtho Hiyamenal *If the restriction is made, then it would not work properly in cases where there is l? 6ne single genitive case suffix but several things to be qualified. [In the string of rules and recurring items] angasya (A 6.4.1) halah (A 6.4.2) anah (A 6.3.111) samprasdranasya (A 6.3.139), hal is to be qualified, an is to be qualified, and samprasdrana is to be qualified [by the single genitive angasya]. But if there is no Restriction, then do as you please in order to qualify several things by means of one single genitive case suffix. As for example: Devadatta's son, hand, blanket. Therefore there is no point in the restriction. ^®The use of the plural here is a bit odd because it only goes to one rule

212

Indian semantic analysis

Patanjali's argument here is that there are cases where one single genitive case suffix serves to denote several relations that may differ according to circumstances. A restriction which limits a genitive case suffix to denote just one particular relation would accordingly create problems. Or, as Kaiyata puts it, one can no longer have the relation between qualifier and qualificand. To illustrate his point, Patanjali adduces the example of A 6.4.1 angasya which contains a genitive to be 'interpreted according to context. The word angasya is considered to recur in A 6.4.2 halah. Recurring there are also the ' items dirghah and anah from A 6.3.111 dhralope purvasya dirgho 'nah,; together with A 6.3.139 samprasdranasya. Rule 6.4.2 halah is thus to be inter-1 preted in accordance with the string angasya halo ynah samprasdranasya , dirghah, 'when preceded by a consonant which is part of a presuffixal base, a j vowel aN {a, i, u) that is samprasdrana (7, w, r, / as replacements of the corre-1 sponding semivowels y, v, r, /) is replaced by a long vowel [when the former f j is the final]81 of a presuffixal base'. For example, rule 6.4.2 halah serves t o | derive the past participle huta from hvd Qive) 'call; invoke', but prevents thef lengthening of u in uta and niruta derived from ve 'weave', in uta because there | is no preceding consonant that is part of the presuffixal base, in niruta because! the r is not a part of the presuffixal base but belongs to the preverb. f Kaiyata clarifies the point in the following manner (MbhP I:411b): | yadlr// 'angasya' iti sthdnasasthitvdd visesanavisesyabhdvo nd prakalpata ityi arthahll | \ 'If [a restriction is made]': that is to say, if [the genitive met with in A 6.4.1] angasyg^ is accepted as a sthdnasasthT, then a relation between qualifier and qualificand is| rendered impossible. This interpretation is supported by Nagesa (MbhU 1:41 lb): tattvdd dhala ity ddind 'vayavavdcakena visesanavisesyabhdvo na sydd ity arthaH/M That is to say, in reality there can be no qualifier-qualificand relatiorj with halah etc| since these refer to a part. "" To get rid of this difficulty, Patanjali simply suggests that if no restriction i] made, one single genitive case suffix may serve to qualify different things i desired, thus denoting various specific relations as in the example 'Devadatta'^ son, hand, blanket' where relations like father-son, part-whole, owner-ownec; are involved. Therefore, even though a genitive case suffix strictly speakin| merely denotes relation in general, the restriction is rendered unnecessary. T| this Patanjali raises the following objection (Mbh 1:119,19-20): nanu coktam ekasatam sasthyarthd ydvanto vd te sarve sasthyam uccdritaya^ prdpnuvantil \
81

A 1.1.52 alo 'ntyasya directs the substitution to the final sound of the form given in the \ tive case; cf. Nagesa's remarks on the Bhasya passage under discussion (MbhU 1:41 lb) < the Kasikd on rule A 6.4.2.

Substitution

213

Well, but it has already been stated that there are one hundred meanings [or, purposes of use] for the genitive case suffix82 - or as many as there are - [and] all of them would apply when a genitive case ending is pronounced. This objection is immediately met with a counter-objection (Mbh 1:119,20-5): naisa dosah I yady api loke bahavo 'bhisambandha drthd yaund maukhdh srauvds ca sabdasya tu sabdena ko 'nyo 'bhisambandho bhavitum arhaty anyad atah sthdndtl sabdasydpi sabdendnantarddayo 'bhisambandhdhf aster bhur bhavatiti samdehah sthdne 'nantare samipa itil samdehamdtram etad bhavati sarvasamdehesu cedam upatisthate vydkhydnato visesapratipattir na hi samdehdd alaksanam itil sthdna iti vydkhydsydmah II There is no fault here. Although there are numerous relations83 in the world - [such as] possessive,84 genetic, oral,85 ritual86 - still, between one linguistic unit and another linguistic unit, what relation is allowed to obtain other than this [one defined by the] sthdnal Even between one linguistic unit and another linguistic unit there are relations such as 'adjacent' etc. So there is doubt when it is stated [in rule 2.4.52 aster bhiih] that bhu occurs 'of as': in the sthdna [of as], in the adjacency, in the nearness? - This is mere doubt, and in all cases of doubt this [paribhdsd] presents itself: 'the understanding of something particular [among two or more possibilities is ascertained] by reasoned explanation, for [a grammatical rule should] not [be considered] a non-rule as a result of [mere] doubt [as to its interpretation].' So we shall interpret [the genitive to denote a relation defined by the locative word form] sthdne. Pataiijali's first point here is that although several different relations can be observed in the world, relations which in language would be expressed by means of a genitive case suffix, only such relations as may possibly obtain between one linguistic unit and another are of interest here. And what particular relation is allowed to obtain between two such terms other than the one defined by the sthdnal In replying to this rhetorical question, Patanjali is forced to admit that even between one linguistic unit and another linguistic unit relations such as dnantarya 'adjacency' may also obtain. From the rest of the passage it is evident that Patanjali, in addition to dnantarya, alludes to the particular relation of sdmipya 'nearness' here. Whether he has even other relations f2 For the compound sasthyarthdh, see pp. 197 ff. above. & The terms sambandha and abhisambandha have been discussed by A. Aklujkar who concludes that 'no technical or grammatical distinction of any kind exists between sambandha and abhisambandha? (1989:303). He goes on: The appearance of, and to some extent the preference for, abhisambandha when its equally non-technical colleague sambandha can convey its 1 meaning seems to be due to the sensitivity which early Sanskrit authors had for the shades of meaning expressed by the upasargas or prefixes. As the situation was probably perceived as one in which word "X" turned to thing "x" or word "Y" for effecting a connection, it was ^perhaps felt that an addition of the prefix abhi, which indicates "facing" or "looking in the 84 direction of", was appropriate.' That is, the relation between owner and owned. P6 That is, the relation between teacher and pupil. ^ That is, the relations between priests in a sacrifice, between the yajamdna and the priests, between mortals and immortals, etc. See Malamoud 1976:155 f.

274

Indian semantic analysis

in mind is hard to tell. The tradition does not mention other reldtions in this context. The early eighteenth-century author Vasudeva Diksita categorically limits them to three in his Bdlamanoramd (SK 1:44,4-5): sabdasya sabdena traya eva sambandhdh - dnantaryam sdmipyam prasangas ca, 'only three relations [may obtain] between a linguistic unit and [another] linguistic unit: adjacency, nearness and possible appearance'.87 Anyhow, it is evident that at least the dnantarya and sdmipya relations may obtain in addition to the one defined by the sthdna. Hence, Patanjali argues, doubt will arise as to the specific sense of a rule such as A 2.4.52 aster bhuh, the purpose of which is to teach that bhu replaces as in the domain of drdhadhdtuka suffixes. The word asteh in this rule exhibits a sthdnasasthi, but how can we be certain about that when also such relations as dnantarya, adjacency, and sdmipya, nearness, may obtain? Because, says Patanjali, this is mere doubt, and in all cases of doubt the following paribhdsd, general principle of interpretation, presents itself: vydkhydnato visesapratipattir na hi samdehdd alaksanam** In a case of doubt, this paribhdsd tells us, we have to ascertain a particular conclusion by reasoned explanation, vydkhydna}9 For a rule in grammar (laksana), which is a means of understanding, does not cease to be so out of mere doubt. Whereas the word alaksana in the present context may be taken in the narrow sense of 'non-rule', the seventh century author Nllakantha glosses it abodhaka 'something which does not give understanding', resorting to the familiar nirvacana of the word laksana: laksyate yneneti laksanamjfidnasddhanam,90 'laksana [is so called according to the analysis] "something is recognised/characterised by means of it", that is, it is a means of knowledge'. This brings us closer to the coiiimon meaning of laksana, namely 'sign;* characteristic mark' or, more precisely, 'that through which something is characterised', thus comprising both definiens and definitio of Western terminology.91 As in my translation of the Bhdsya passage above, the paribhdsd could] thus be interpreted so as to have a wider scope than just the^rules of the; AstddhydyT. Invoking this paribhdsd, Patanjali explicitly says that we are fbrced to inter-j pret. And, he continues, the specific relation which the genitive9 marks in thg Note that dnantarya, sdmipya and prasanga are called relations here. The word sthdna is takenl in the sense of prasanga 'possible appearance'. ^ 88 This way of arguing (samdehamdtram etc.) is met with in the form of a fixed expression af several places in the Bhdsya, for example Mbh 1:42,11 (A 1.2.2); 266,15 (A 1.3.9); 271,15 (Ai 1.3.10); II: 106,3 (A 3.2.57); 106,18 (A 3.2.57). The paribhdsd is also invoked on its own, fol example Mbh 1:6,26; 7,5 (Paspasd); 111:246,3 (A 7.1.13), and is widely accepted by Paninlya| from Patanjali onwards. Nagesa presents it as the very first paribhdsd in hi| Paribhdsendusekhara and it is listed in the majority of grammatical paribhdsd works (cf| Abhyankar 1967, index). 89 According to Nagesa: vydkhydndc chistakrtdt, 'from the interpretation done by the sistas, thgf learned' (Paribhdsendusekhara, ed. Abhyankar 1962:1); cf. Thieme 1957:53 ff. 90 Nllakantha's Paribhdsdvrtti, ed. Abhyankar 1967:293. Patanjali resorts to the .same nirvacan^. in his Mahdbhdsya on A 2.3.37, stating that a laksana is that by which something is recojil 91 msed (Mbh 1:459,2-3; 7-10). See G. Gren-Eklund 1984:118, note 53.\ i
87

Substitution

215

present context we shall interpret by supplying the locative word form sthdne. The commentators take us considerably further here, providing reasons for the choice of this particular interpretation. Kaiyata remarks (MbhP I:412a): sthana itil antarangatvdd laksydpeksandd veti bhdvahl [Why would we understand] 'sthdne' [?] The idea is: either from the fact that it is present first in the mind {antaranga), or from considering the laksya, the final result of an operation. Kaiyata identifies two reasons why we should interpret a genitive by supplying the locative word form sthdne. On the one hand, such an interpretation is ascertained through the principle of antarangatva. In brief, an operation92 is considered antaranga when its conditions or causes of application are internal or proximate with regard to those of another operation, accordingly termed bahiranga, whose conditions are external.93 Moreover, what depends on a meaning that presents itself prior to another meaning is also termed antaranga. In the process of formation an antaranga operation is present in the mind prior to a bahiranga operation, and is thus considered stronger than it.94 In other words, that which presents itself earlier takes effect earlier. This idea is reflected in the widely accepted paribhdsd: asiddham bahirangam antarange, 'a bahiranga [operation] is not effected when [there 4S a possibility of] an antaranga [operation taking effect]'.95 Patanjali himself accepts the principle of antarangatva and quotes this paribhdsd at several places in the Mahdbhdsya though he is occasionally forced to make exceptions to it.96; On the force of this antara/iga-principle, then, it is possible for Kaiyata to claim that the relation marked by a genitive should be interpreted by supplying the locative woroNform sthdne. This is what comes to mind first.97 On the other hand\]jCaiyata says, this interpretation is arrived at also by igksydpeksana, that is, by paying attention to the final forms which are the output of grammatical operations. The interpretation whereby one supplies the word form sthdne is the only one which would give the forms desired, the forms met with in actual Sanskrit usage.
92

I take 'operation' here to include a rule as well as what is taught in a rule. ^ See p. 196, note 44 above. 91 Note here the following remark of Joshi and Roodbergen (1980:ii). With regard to the view that the whole sentence, and not only the single words of which it is composed, is at the basis of formation, they state that 'a distinction must be introduced between antaranga-m\es which are word-integrative rules, and bahiranga-rules which are non-word-integrative rules. Moreover, as regards the order of application, priority must be given to the antaranga-mles.' ^Kagesa presents this as paribhdsd 50 in his Paribhdsendusekhara. It is listed in all extant works on grammatical paribhdsds. For recent critical work on this paribhdsd, see Kiparsky 1982:87 ff. and Bronkhorst 1986. Even though the antaranga-paribhdsd may turn out to be irrelevant as Kiparsky (1982) has tried to demonstrate, it was certainly at work in the minds of Sanskrit authors at least from the times of Patanjali. m For example, Mbh 1:26,14 (Sivasutra 4); 111:93,3 (A 6.1.135); 347,22 (A 7.4.10). m This will be explained below.

276

Indian semantic analysis

Bhattoji Dlksita retains the arguments based on laksydpeksana and antarangatva, but adds a third one in his Sabdakaustubha (SK 244,10-12): aster bhur itydddv anekasambamdhaprasamge 'pi laksydnurodhdd antaramgatvdt sthdnivad ddesa itijndpakdc ca sthdneyogataiva vydkhydsyate I Even though several relations are possible with regard to [a rule] such as [A 2.4.52] aster bhuh, only the interpretation 'sthdneyogd' should be offered, because it complies with the final result of an operation, because it is antarahga [with regard to any other relation], and, moreover, because [rule A 1.1.56] sthdnivad ddeso ['nalvidhau] serves as an indicator98 [in this respect]. Bhattoji says the relation defined by the expression sthdneyogd comes to mind first, and it is the only one that fulfils the demands of the forms actually met with in the real language. Moreover, he argues, rule 1.1.56 sthdnivad ddeso 'nalvidhau serves as an indication for this interpretation of rule 1.1.49 as applied to a rule such as A 2.4.52 aster bhuh. As mentioned already, rule 1.1.56 sthdnivad ddeso 'nalvidhau teaches that, with specific exceptions, a substitute, ddesa, is considered like the sthdnin, the element which undergoes the substitution in an operation. Nagesa, commenting on Kaiyata's statement, pays attention^ only to the argument based on antarangatva but retains Bhattoji's argument that A 1.1.56 is a jndpaka here. Nagesa's remarks are important since he "claims that antarangatva is established regardless of whether we interpret the term sthdna as artha 'meaning' or prasanga 'possible appearance' 99 (MbhU I:412a): antarangatvad itil sthdnasabdo 'rthavdci prasangavdcT vd, ubhayathd ypy antarangatvam I sthdnivad ddesa itijndpakam apy atra bodhyamll 'From the fact that it is present first in the mind {antarangd)? The word sthdna signifies artha 'meaning', or prasanga 'possible appearance', but! either way antarangatva [is established]. Moreover, [A 1.1.56] sthdnivad ddeso 'nalvidhau should be recognised as an indicator here. Kaiyata's claim that we would supply the locative word form sthane simply on the basis of antarangatva is thus clearly supported by Nagesa wiio adds that however we interpret the word sthdna, we have antarangatva. Vaidyanatha Payagunda, Nagesa's pupil, justifies how antarangatva is established regardless of which of the two senses we ascribe to the word sthdna. In his Chdyd (Bhargavasastrl JosI 1951:412, note 4) he comments on Nagesa's statement as follows: The technical term jndpaka may best be rendered 'something which serves to indicate'. Based on supposed implications which result from internal analysis of the rules ^f grammar, a jndpaka is a structural argument through which a valid interpretation can be inferred and justified. In other words, when seemingly conflicting features occur in the grammar, one should look for some other feature which indicates the valid interpretation on the basis of consistency and a unified system. i 99 These are the two senses most commonly attributed to the term sthdna by the tradition. They will be discussed in detail below.
98

Substitution

217

ddye 'bhidheyasya sabdaprayogahetutvenddau buddhisamnidhdndt, antye prasaktasyaivdnantaryddisambhavena tasyddau taddrudhattvam [sic] ity arthah// anyathd sutravaiyarthyam apity aha - sthdnill On the first view, [that sthdna signifies artha,] because that which is to be denoted is present in the mind first since it is the reason for using words; on the second view, [that sthdna signifies prasanga,] since the possibility of [relations] such as adjacency exists only for what would possibly appear (prasaktasya), the latter reaches it (=the mind) at the very beginning; that is the sense [of Nagesa's remark]. Moreover, since otherwise the rule [A 1.1.49] is meaningless, he says: sthdnivat etc. Vaidyanatha's argument is as follows. On the view that the term sthdna signifies artha, meaning, the antarangatva of the srtena-relation is justified through the fact that what is to be denoted or expressed is the cause for the use of a word, and accordingly that is established first in the mind/That is to say, meaning comes first when we use words and therefore one can claim that there is antarangatva. But if it is assumed that sthdna signifies artha, then one does not need Panini's rule, since words are used to denote or express artha and that is in the mind already. On the view that sthdna signifies prasanga, the possible appearance of a linguistic unit, we can also interpret the genitive with the supplement of the locative word form sthdne even without rule 1.1.49. The objection would be that there are also relations such as adjacency and nearness. The answer to this is that sthdna here signifies prasanga 'possible appearance'. The possibility of these other relations applying exists only for what is prasakta, for what would possibly appear. In other words, there exist relations such as adjacency and nearness, but these can obtain only for something which is about to appear. By way of example, A 2.4.52 aster bhuh would then be interpreted to mean that since the possible appearance of as is antaranga, bhu is used when there is a u possible appearance of as. To state that 'bhu is used next to as' would not be possible, since adjacency or nearness is possible only for something which will ; appear. These relations presuppose whatjs-going to be there. Hence the ! sthdna-vtl&iion, which applies to^what is prasakta, to what would appear, is 1 what one arrives at first, and the st/iana-relation is thus antaranga with regard to any other relation. Only something which is prasakta can be a relatum in the dnantarya and sdmipya relations. These hold only between word and word, not between word I and meaning. This is of significance with regard to the two senses of sthdna, I namely artha and prasanga, referred to by Nagesa. Either way antarangatva is ; established. The alternative senses of artha and prasanga are just choices with regard to the weight being laid on meaning or on word form. £ Returning to the Bhdsya discussion, it is now possible to see the full significance of Patanjali's conclusion: sthdna Hi vydkhydsydmah, 'so we shall interpret [the genitive to denote a relation defined by the locative word form] osthdne', which in fact is his last statement on A 1.1.49 itself. From the justifica:tion of this interpretation by the principle of antarangatva resorted to by the

218

Indian semantic analysis

commentators, one may actually conclude that rule 1.1.49 is superfluous. And the necessity of the rule is indeed what Patanjali rhetorically questions in his final remarks on A 1.1.49 (Mbh 1:119,25-8): na tarhiddnim ayam yogo vaktavyah I vaktavyas ca I kim prayojanaml sasthyantam sthdnena yathd yujyeta yatah sasthy uccdntdl kim krtam bhavdtil nirdisyamdnasyddesd bhavantity esd paribhdsd na kartavyd bhavatill In that case, then, this rule need not be stated. It must be stated. For what purpose? In order that what ends in a genitive case suffix, provided the genitive is actually pronounced after it, should be linked with the sthdna. What is accomplished by this? The paribhdsd which states that 'substitutes occur in the place of what is expressly enunciated' need not be made. Rejecting at first any reason for stating the rule, Patanjali then says it should be stated. What is taught by the rule may be fully provided for by the principle of antarahgatva. But the rule serves another purpose, namely that what ends in a genitive, provided the genitive case suffix is actually pronounced after it, should be linked with or, possibly through the sthdna. Interpreted in this way, the rule is made to render the paribhdsd that states: nirdisyamdnasyddesd bhavanti, 'substitutes occur in the place f of what is expressly enunciated'. Bhattoji Dlksita sums the situation up as follows in his Sabdakaustubha (SK 244,12-15): tasmdn ndyam etasya sutrasya sutrdrthahl kim tu sasthyamtam ucbdryamdnam eva sthdnena yujyate na tu pratvyamdnam iti sutrdrthahl tad etad ucyate nirdisyamdnasyddesd bhavamtiti I Therefore this is not the way to understand this rule.100 No, the meaning of the rule is: 'only that which is expressly enunciated with a genitive case ending is linked with the sthdna, not that which is [merely] understood'. So what is [really] stated is [the paribhdsd]: 'substitutes occur in the place of what is expressly enunciated'. The paribhdsd is not rejected, it simply does not require special statement. In fact, there are numerous instances where Patanjali resorts to it when discussing operations that could be misapplied. The paribhdsd is thus cited at several places in the Mahdbhdsya, most notably at A 6AA30pddahpat where Patanjali deals with it at some length.101 In his third vdrttika on this rule Katyayana refers to it in order to meet a difficulty raised by the second vdrttika.102 At this point Patanjali proclaims the necessity of the paribhdsd, telling
100

101

102

That is, to identify the genitive in a rule such as 2.4.52 aster bhilh as a sthdnasdsthTsince this would come about anyway, either because such an interpretation is antaranga, or because laksydpeksana would bring it about, or because A 1.1.56 would serve to indicate it. Other instances are: Mbh 1:29,23 (Sivasutra 5); 116,19 (A 1.1.47); 129,16 (A 1.1.51), 111:20,26 (A 6.1.13); 22,10 (A 6.1.13); 50,18 (A 6.1.71); 186,1 (A 6.4.19); 214,1 (A 6.4.101); 225,11 (A 6.4.142); 236,4 (A 7.1.1); 252,20 (A 7.1.33); 258,5 (A 7.1.50); 398,6 (A 8 2.16)'. A 6.4.130, vt. 3: na vd nirdisyamdnasyddesatvdt, 'or rather not, because substitutes [occur] in the place of what is expressly enunciated'.

Substitution

219

us that there are numerous purposes served by it.103 These are listed in the remaining vdrttikas and thoroughly discussed by Patanjali who demonstrates how errors would come about if one does not accept Xhe paribhdsd. His conclusions here are however consistent with his final remarks on A 1.1.49. The functions served by the paribhdsd are in fact taken care of by rule 1.1.49 sasthT sthdneyogd (Mbh III:222,16-17):104 sd tarhy esd paribhdsd kartavydl na kartavydl uktam sasthT sthdneyogety etasya yogasya vacane prayojanam sasthyantam sthdnena yathd yujyeta yatah sasthy uccdritetill So then this paribhdsd needs to be stated. It need not be. It has been stated that the purpose in formulating rule [A 1.1.49] sasthT sthdneyogd is that whatever ends in a genitive case suffix, provided the genitive is actually pronounced after it, should be linked with the sthdna. If rule 1.1.49 is interpreted so as to render the paribhdsd, then the scope of the rule and the scope of the paribhdsd are logically interconnected. Since Panini does not himself explicitly state the paribhdsd, it seems in the first place exigent to consider the reasons given by the tradition for introducing the paribhdsd and see whether these harmonise with the rules and principles formulated in the Astddhydyl In doing so it is necessary to look at the arguments adduced by Paninlyas in the context of this nirdisyamdna-principle. What Paninlyas remark on the paribhdsd may thus also contribute to a better understanding of how rule 1.1.49 and the operations governed by it have been viewed by the tradition. / The nirdisyamana-paribhasa / Invoked by Paninlyas from Katyayana onwards, the nirdisyamdna-paribhdsd is listed in the majority of extant grammatical paribhdsd collections.105 Nagesa presents it as paribhdsd 12 in his Paribhdsendusekhara. Aiming at consistency in the body of rules and the principles which govern their interpretation, Nagesa is first and foremost concerned with the question of how the paribhdsd interacts with other rules of grammar. Hence, his discussion provides us with a valuable structural frame for presenting the problems which according to the tradition pertain to the present paribhdsd. With this in mind, it seems more rewarding to review the discussion in the Paribhdsendusekhara than to investigate Patanjali's remarks on the paribhdsd at rule 6.4.130 directly since most of the issues raised in the Mahdbhdsya are also dealt with in the course of Nagesa's exposition. So are the issues brought up by Kaiyata and by Nagesa himself in the context of Patanjali's final remarks at rule 1.1.49. These are also
103 104 105

Mbh 111:221,11: avasyam esd paribhdsd kartavydl bahuny etasydh panbhdsdydh prayojandm I The same view is put forward also in the paribhdsd works of Purusottamadeva, Slradeva, Nllakantha and Haribhaskara (ed. Abhyankar 1967:127, 167, 294, 320). See Abhyankar 1967, index.

220

Indian semantic analysis

more easily understood in the light of Nagesa's exposition in the Paribhdsendusekhara and will accordingly be considered only after an examination of the paribhdsd has been carried out. Nagesa's exposition is, however, extremely dense. To present a paraphrase and an explanation of his arguments seems therefore a more favourable way to proceed here than to give a literal translation of the whole discussion.106 The trick here in the first place is getting the relations right between the paribhdsd and certain rules with which it seems to be in conflict.. Basically, these are the rules A 1.1.52 alo 'ntyasya, A 1.1.54 ddeh parasya, A 1.1.55 anekdlsit sarvasya and A 1.1.72 yena vidhis tadantasya, and, moreover, another paribhdsd not listed as a separate rule in the Astddhydyi: yaddgamds tadgumbhutds tadgrahanena grhyante. Rule 1.1.52 alo 'ntyasya states that in replacements of sounds, the substitution applies to the final sound of the element which undergoes substitution, that is, to the final sound of the form cited in the genitive case in a rule. As an exception to this, rule 1.1.54 ddeh parasya states that when an operation is taught for a following form, the substitution is directed to the initial sound of that form. Another exception is taught by rule 1.1.55 anekdlsit sarvasya which teaches that a substitute marked with a diacritic S or one consisting of more than one sound replaces the entire element which undergoes substitution, instead of replacing only its final sound. Rule 1.1.72 yena vidhis tadantasya stipulates that when an operation is taught by means of some linguistic unit, that unit denotes, apart from itself, also whatever ends with it. Finally, there is the general principle of interpretation yaddgamds tadgumbhutds tadgrahanena grhyante which states that augments which are added to some element, thus forming an integrated part of it, are understood when this element is mentioned.107 Nagesa presents the latter paribhdsd, referred to in what follows as theyadagama-principle, zsparibhdsd 11 in his Paribhdsendusekhara, that is, immediately before the' paribhdsd which is the present object of investigation. Nagesa introduces the nirdisyamdna-paribhdsd^as he usually does any paribhdsd included in his work, by pointing to a difficulty which has to be met. In the present case the difficulty is caused right away by the preceding paribhdsd, the yaddgama-princvple. The problem is that the property of being a substituend, sthdnitva, does not apply correctly to linguistic units recognised by the yaddgama-phnciple. With regard to an aorist formation such as udasthdt, we have to consider the following situation. Rule A 8.4.61 udah sthdstambhoh purvasya teaches the replacement by a sound which is homogeneous with the preceding for sthd and stambh when these follow the preverb ud.m Since here a replacement is taught
106 107 108

A full translation has been made by Kielhorn (1871-^.67 ff.). Katyayana refers to this paribhdsd in vdrttika 5 on rule A 11 20 dddhd ghv addp. This rule accounts for forms such as the infinitive utthdtum. By rule 8.4.65 jharo jhan savarne, a t which is preceded by a consonant and followed by a homogeneous sound such as th is elided

Substitution

221

for a following form, rule 1.1.54 ddehparasya applies and directs the substitution to the initial sounds of sthd and stambh.109 Now, the yaddgama-princiiple states that an element to which an augment is added denotes not only itself, but also that which results from its combination with that augment. In the formation of the aorist form udasthdt, the augment aT, acutely accented a, is added to sthd. Accordingly, sthd denotes also asthd which in the formation of udasthdt is preceded by ud. In the sound-replacement taught by rule A 8.4.61, then, the augment a would be identified as the sthdnin, the element which is to undergo the substitution operation. In order to avoid such an erroneous occurrence of homogeneity with the preceding sound, paribhdsd 12 is introduced: nirdisyamdnasyddesd bhavanti, 'substitutes occur in the place of what is expressly enunciated'. The paribhdsd thus excludes the possibility of ascribing sthdnitva, the property of being a substituend, to elements which are recognised by application of the yaddgamaprinciple. Having thus given a reason for introducing the paribhdsd, Nagesa goes on to remark on its relationship with rule 1.1.49 sasthT sthdneyogd. Since the general sense of this rule, that the genitive marks the particular relation in some way defined by the locative word form sthdne, is evident from antarangatva or laksydpeksana, Nagesa claims that the rule is read twice. And the paribhdsd conveys the sense of the second reading: only something which is expressly enunciated with a genitive case ending, or, more precisely, only that which in the real language is like that which is enunciated or forms part of110 what is enunciated, becomes a term in the relation defined by the sthdna, not that which is merely understood to be so. In other words, this paribhdsd prevents a linguistic unit which is merely inferred and not expressly enunciated with a genitive case ending from being recognised as a sthdnin, an element which is to undergo a substitution operation. Such a recognition could, as noted, erroneously result from the application of the yaddgama-principle formulated by paribhdsd 11. Therefore, Nagesa says, the nirdisyamdna-paribhdsd is established. The validity of this paribhdsd is then challenged by a series of objections which have to be met. The first of these concerns operations taught in a rule such as A 7.4.32 asya cvau. This rule teaches that a final a or d of a presuffixal base is replaced by f when the suffix Cvlm follows. Only short a is expressly enunciated in the rule. On the view that a linguistic unit denotes the generic character which is inherent in every individual unit, Nagesa argues that there
109

110 111

Concentrating on the problem raised by the yaddgama-pnnciple, Nagesa leaves rule 11 54 ddehparasya out of his discussion here, probably on purpose, since the rule later on is shown to apply together with the nirdisyamdna-panbhdsd, thus directing the substitution to the initial sound of what is actually enunciated in a rule with a genitive case ending. Seep 228below The adverbial suffix Cvl (A 5 4.50 krbhvastiyoge sampadyakartan cvih) is deleted by A 6 1.67 ver aprktasya. Rule A 7 4.32 accounts for forms such as suklfkrta 'made white' where the final a of sukla 'white' is replaced by F.

222
i s n o f a u l

Indian semantic analysis
• * » t h e r e , s i n c e a n e s s i n h e r e s a l s o i n l o n g a s o t h a t w h e n s h o r t a i s m e n -

t

i

o

n

e

d

i

n

a

r

u

l

e

,

l

o

n

g

a

i

s

a

l

s

o

d

e

n

o

t

e

d

.

1

1

2

T

o

b

u

t

t

r

e

s

s

h

i

s

a

r

g

u

m

e

n

t

,

N

a

g

e

s

a

a

l

s

o

r

e

f

e

r

s

t

o

t

h

e

p

r

o

h

i

b

i

t

i

o

n

t

a

u

g

h

t

b

y

A

6

.

4

.

8

5

n

a

b

h

u

s

u

d

h

i

y

o

h

.

A

n

o

p

e

r

a

t

i

o

n

o

n

a

s

o

u

n

d

e

x

p

r

e

s

s

l

y

e

n

u

n

c

i

a

t

e

d

i

n

a

r

u

l

e

a

p

p

l

i

e

s

t

o

a

l

l

v

a

r

i

e

t

i

e

s

d

e

n

o

t

e

d

b

y

t

h

a

t

p

a

r

t

i

c

u

l

a

r

s

o

u

n

d

a

c

c

o

r

d

i

n

g

t

o

r

u

l

e

A

1

.

1

.

6

9

a

n

u

d

i

t

s

a

v

a

r

n

a

s

y

a

c

d

p

r

a

t

y

a

y

a

h

w

h

i

c

h

s

t

i

p

u

l

a

t

e

s

t

h

a

t

a

s

h

o

r

t

v

o

w

e

l

a

l

s

o

d

e

n

o

t

e

s

a

n

y

v

a

r

i

e

t

y

o

f

t

h

a

t

V

o

w

e

l

,

t

h

u

s

a

l

s

o

i

t

s

c

o

r

r

e

s

p

o

n

d

i

n

g

l

o

n

g

,

p

r

o

v

i

d

e

d

t

h

e

v

o

w

e

l

i

n

q

u

e

s

t

i

o

n

i

s

n

o

t

a

s

u

f

f

i

x

.

R

u

l

e

6

.

4

.

8

5

p

r

o

h

i

b

i

t

s

p

r

e

v

o

c

a

l

i

c

s

e

m

i

v

o

w

e

l

s

u

b

s

t

i

t

u

t

i

o

n

i

n

t

h

e

c

a

s

e

o

f

a

p

r

e

s

u

f

f

i

x

a

l

b

a

s

e

e

n

d

i

n

g

i

n

b

h

u

o

r

s

u

d

h

l

B

o

t

h

o

f

t

h

e

s

e

e

n

d

i

n

a

l

o

n

g

v

o

w

e

l

.

N

o

v

y

-

,

i

f

t

h

e

p

r

e

-

v

o

c

a

l

i

c

s

e

m

i

v

o

w

e

l

s

u

b

s

t

i

t

u

t

i

o

n

p

r

o

v

i

d

e

d

f

o

r

b

y

t

h

e

p

r

e

c

e

d

i

n

g

r

u

j

e

s

A

6

.

4

.

8

2

e

r

a

n

e

k

d

c

o

'

s

a

m

y

o

g

a

p

u

r

v

a

s

y

a

a

n

d

A

6

.

4

.

8

3

o

h

s

u

p

i

d

i

d

n

o

t

a

p

p

l

y

a

l

s

o

t

o

t

h

e

l

o

n

g

v

o

w

e

l

s

f

a

n

d

w

,

t

h

e

n

t

h

e

p

r

o

h

i

b

i

t

i

o

n

t

a

u

g

h

t

b

y

r

u

l

e

6

.

4

.

8

5

w

e

f

u

l

d

b

e

o

u

t

o

f

p

l

a

c

e

.

T

h

e

s

e

c

o

n

d

p

o

s

s

i

b

l

e

o

b

j

e

c

t

i

o

n

t

o

t

h

e

p

a

r

i

b

h

d

s

d

c

o

n

c

e

r

n

s

t

h

e

a

n

u

b

a

n

d

h

a

,

d

i

a

c

r

i

t

i

c

m

a

r

k

e

r

,

N

w

h

i

c

h

s

e

r

v

e

s

,

a

m

o

n

g

o

t

h

e

r

f

u

n

c

t

i

o

n

s

,

t

o

m

a

r

k

s

u

b

s

t

i

t

u

t

e

s

a

s

a

p

p

l

y

i

n

g

o

n

l

y

t

o

t

h

e

f

i

n

a

l

s

o

u

n

d

o

f

t

h

e

e

l

e

m

e

n

t

w

h

i

c

h

u

n

d

e

r

g

o

e

s

s

u

b

s

t

i

t

u

-

t

i

o

n

.

I

t

m

a

y

b

e

a

r

g

u

e

d

t

h

a

t

t

h

e

n

i

r

d

i

s

y

a

m

d

n

a

-

p

a

r

i

b

h

d

s

d

m

a

k

e

s

t

h

i

s

m

a

r

k

e

r

N

s

u

p

e

r

f

l

u

o

u

s

i

n

r

u

l

e

A

6

.

4

.

7

7

a

c

i

s

n

u

d

h

d

t

u

b

h

r

u

v

d

m

y

v

o

r

i

y

a

n

u

v

a

n

a

u

.

T

h

i

s

r

u

l

e

t

e

a

c

h

e

s

t

h

a

t

w

h

e

n

a

v

o

w

e

l

f

o

l

l

o

w

s

,

t

h

e

i

a

n

d

w

,

b

o

t

h

s

h

o

r

t

a

n

d

l

o

n

g

,

w

h

i

c

h

a

r

e

f

i

n

a

l

i

n

a

p

r

e

s

u

f

f

i

x

a

l

b

a

s

e

w

h

i

c

h

e

n

d

s

i

n

t

h

e

c

o

n

j

u

g

a

t

i

o

n

a

l

c

l

a

s

s

s

u

f

f

i

x

S

n

u

,

o

r

i

n

a

r

o

o

t

,

o

r

b

h

r

u

'

b

r

o

w

'

a

r

e

r

e

p

l

a

c

e

d

r

e

s

p

e

c

t

i

v

e

l

y

b

y

i

y

A

N

a

n

d

u

v

A

N

.

T

h

e

d

i

a

-

c

r

i

t

i

c

m

a

r

k

e

r

N

s

i

g

n

a

l

s

t

h

a

t

t

h

e

s

u

b

s

t

i

t

u

t

i

o

n

a

f

f

e

c

t

s

o

n

l

y

t

h

e

f

i

n

a

l

s

o

u

n

d

o

f

t

h

e

e

l

e

m

e

n

t

s

l

i

s

t

e

d

.

B

u

t

/

a

n

d

u

h

a

v

e

b

e

e

n

e

x

p

r

e

s

s

l

y

e

n

u

n

c

i

a

t

e

d

i

n

t

h

e

r

u

l

e

b

y

y

v

o

h

.

H

e

n

c

e

,

i

f

t

h

e

p

a

r

i

b

h

d

s

d

i

s

v

a

l

i

d

t

h

e

s

u

b

s

t

i

t

u

t

i

o

n

w

o

u

l

d

a

f

f

e

c

t

o

n

l

y

t

h

e

s

o

u

n

d

s

i

a

n

d

u

e

v

e

n

i

f

t

h

e

s

u

b

s

t

i

t

u

t

e

s

i

y

a

n

d

u

v

a

r

e

n

o

t

m

a

r

k

e

d

w

i

t

h

a

d

i

a

c

r

i

t

i

c

N

w

h

i

c

h

t

h

e

r

e

f

o

r

e

b

e

c

o

m

e

s

s

u

p

e

r

f

l

u

o

u

s

.

H

o

w

e

v

e

r

,

N

a

g

e

s

a

a

r

g

u

e

s

,

t

h

i

s

c

l

a

i

m

l

o

s

e

s

i

t

s

f

o

r

c

e

b

e

c

a

u

s

e

'

r

o

o

t

s

'

e

t

c

.

h

a

v

e

a

l

s

o

b

e

e

n

e

x

p

r

e

s

s

l

y

e

n

u

n

c

i

a

t

e

d

i

n

t

h

e

r

u

l

e

,

a

n

d

s

o

w

e

d

o

h

a

v

e

t

o

a

t

t

a

c

h

t

h

e

a

n

u

b

a

n

d

h

a

N

t

o

i

y

a

n

d

u

v

t

o

m

a

r

k

t

h

e

m

a

s

s

u

b

-

s

t

i

t

u

t

e

s

o

f

a

f

i

n

a

l

s

o

u

n

d

o

n

l

y

.

\

N

a

g

e

s

a

t

h

e

n

p

u

s

h

e

s

t

h

e

a

r

g

u

n

i

e

n

t

f

u

r

t

h

e

r

,

q

u

e

s

t

i

o

n

i

n

g

t

h

e

"

U

s

e

o

f

t

h

e

a

n

u

b

a

n

d

h

a

N

i

n

r

T

N

a

n

d

r

i

N

.

T

h

e

s

e

a

r

e

s

u

b

s

t

i

t

u

t

e

s

J

S

g

r

t

h

e

s

h

o

r

t

v

d

w

e

l

r

i

n

t

h

e

o

p

e

r

a

t

i

o

n

s

t

a

u

g

h

t

b

y

r

u

l

e

s

A

7

.

4

.

2

7

r

i

n

r

t

a

h

a

n

d

A

7

.

4

.

2

8

r

i

n

s

a

y

f

a

g

l

i

n

k

s

u

.

m

1

1

2

T

h

i

s

v

i

e

w

i

s

k

n

o

w

n

a

s

t

h

e

j

d

t

i

p

a

k

s

a

,

s

t

r

i

c

t

l

y

s

p

e

a

k

i

n

g

t

h

e

v

i

e

w

t

h

a

t

a

n

o

u

n

r

e

f

e

r

s

t

o

a

g

e

n

u

s

,

a

s

o

p

p

o

s

e

d

t

o

t

h

e

v

y

a

k

t

i

p

a

k

s

a

a

c

c

o

r

d

i

n

g

t

o

w

h

i

c

h

a

n

o

u

n

r

e

f

e

r

s

t

o

a

n

i

n

d

i

v

i

d

u

a

l

e

n

t

i

t

y

.

A

p

p

l

i

e

d

t

o

P

a

n

i

n

i

'

s

r

u

l

e

s

,

t

h

e

f

o

r

m

e

r

v

i

e

w

h

o

l

d

s

t

h

a

t

o

n

e

r

u

l

e

c

o

v

e

r

s

a

l

l

p

o

s

s

i

b

l

e

i

n

s

t

a

n

c

e

s

,

w

h

e

r

e

a

s

t

h

e

l

a

t

t

e

r

v

i

e

w

h

o

l

d

s

t

h

a

t

a

r

u

l

e

h

a

s

t

o

b

e

r

e

p

e

a

t

e

d

f

o

r

e

v

e

r

y

i

n

d

i

v

i

d

u

a

l

i

n

s

t

a

n

c

e

.

T

h

i

s

i

s

s

u

e

i

s

c

e

n

t

r

a

l

t

o

t

h

e

M

l

m

a

m

s

a

k

a

s

i

n

t

h

e

i

r

i

n

t

e

r

p

r

e

t

a

t

i

o

n

o

f

V

e

d

i

c

i

n

j

u

n

c

t

i

o

n

s

.

D

o

e

s

a

n

i

n

j

u

n

c

t

i

o

n

s

u

c

h

a

s

b

r

d

h

m

a

n

o

n

a

h

a

n

t

a

v

y

a

h

,

'

a

B

r

a

h

m

a

n

s

h

o

u

l

d

n

o

t

b

e

k

i

l

l

e

d

*

,

r

e

f

e

r

t

o

e

v

e

r

y

B

r

a

h

m

a

n

o

r

t

o

o

n

e

i

n

d

i

v

i

d

u

a

l

?

N

a

g

e

s

a

d

i

s

c

u

s

s

e

s

t

h

e

t

w

o

v

i

e

w

s

a

t

p

a

r

i

b

h

d

s

d

4

0

i

n

h

i

s

P

a

r

i

b

h

d

s

e

n

d

u

s

e

k

h

a

r

a

(

e

d

.

A

b

h

y

a

n

k

a

r

1

9

6

2

:

6

6

f

f

.

;

K

i

e

l

h

o

r

n

1

8

7

1

-

4

:

1

9

0

f

f

.

)

.

1

1

3

R

u

l

e

A

7

.

4

.

2

7

t

e

a

c

h

e

s

t

h

a

t

t

h

e

f

i

n

a

l

o

f

a

p

r

e

s

u

f

f

i

x

a

l

b

a

s

e

i

s

r

e

p

l

a

c

e

d

b

y

r

i

w

h

e

n

'

t

h

e

a

d

v

e

r

b

i

a

l

s

u

f

f

i

x

C

v

l

o

r

a

s

u

f

f

i

x

b

e

g

i

n

n

i

n

g

w

i

t

h

y

a

w

h

i

c

h

i

s

n

o

t

a

k

r

t

o

r

a

s

d

r

v

a

d

h

d

t

u

k

a

s

u

f

f

i

x

f

o

l

l

o

w

s

.

T

h

i

s

a

c

c

o

u

n

t

s

f

o

r

f

o

r

m

s

s

u

c

h

a

s

m

d

t

n

y

a

t

i

'

"

m

o

t

h

e

r

s

"

;

c

o

n

s

i

d

e

r

s

a

m

o

t

h

e

r

'

(

d

e

n

o

m

i

n

a

t

i

v

e

f

r

o

m

m

d

t

r

-

'

m

o

t

h

e

r

'

)

a

n

d

m

d

t

n

b

h

u

t

a

'

b

e

c

o

m

e

a

m

o

t

h

e

r

'

.

R

u

l

e

A

7

.

4

.

2

8

t

e

a

c

h

e

s

.

t

h

a

t

r

i

i

s

s

u

b

-

s

t

i

t

u

t

e

d

f

o

r

t

h

e

f

i

n

a

l

r

o

f

a

r

o

o

t

w

h

i

c

h

i

s

a

p

r

e

s

u

f

f

i

x

a

l

b

a

s

e

b

e

f

o

r

e

t

h

e

c

o

n

j

u

g

a

t

i

o

n

a

l

c

l

a

s

s

s

u

f

f

i

x

S

a

,

t

h

e

p

a

s

s

i

v

e

s

u

f

f

i

x

y

a

K

,

a

n

d

t

h

e

o

p

t

a

t

i

v

e

a

n

d

p

r

e

c

a

t

i

v

e

a

u

g

m

e

n

t

y

d

s

U

T

.

T

h

u

s

,

f

o

r

e

x

a

m

p

l

e

,

t

h

e

r

o

f

V

&

r

'

d

o

;

m

a

k

e

'

i

s

r

e

p

l

a

c

e

d

b

y

r

i

b

e

f

o

r

e

y

a

K

,

g

i

v

i

n

g

t

h

e

3

.

s

g

.

p

a

s

s

i

v

e

f

o

r

m

k

r

i

y

a

t

e

'

i

s

m

a

d

e

'

,

a

n

d

b

e

f

o

r

e

y

d

s

U

T

,

g

i

v

i

n

g

t

h

e

2

.

p

i

.

a

c

t

i

v

e

p

r

e

c

a

t

i

v

e

f

o

r

m

k

r

i

y

d

s

t

a

'

y

o

u

s

h

o

u

l

d

m

a

k

e

'

.

Substitution

223

Here the vowel r alone is expressly enunciated. On the one hand, then, if the nirdisyamdna-paribhdsd is valid, the anubandha N seems to have no function whatsoever, except that it might have been added solely for the sake of clarity. On the other hand, a claim that the anubandha N serves to indicate the paribhdsd is useless with regard to rules where a sound is expressly enunciated conflicts with the Bhdsya on the fifth Sivasutra, hayavarat. One of the problems discussed there is why the ayogavdha sounds114 are not included in the Sivasutras. This prevents them from being referred to by the pratydhdra115 term al which comprises all the other sounds of the Sanskrit language. Discussing possible purposes of including the ayogavdha sounds, Katyayana suggests in vdrttika 8 that they ought to be included, one of the advantages being operations that are prescribed for a final sound. The rule referred to here is A 1.1.52 alo 'ntyasya, 'in the place of the final aV, which directs the substitution to the final sound, referred to by the term al, of the linguistic element which is the target of the substitution operation. Since the ayogavdha sounds are not covered by the term al, they do not fall under the scope of this rule. Hence, one might claim, rule 1.1.52 would be a purpose for teaching them at the stage of the Sivasutras. For example, s replaces the visarjaniya (h) of vrksah when t follows, as in vrksas tatra 'the tree, there', according to A 8.3.34 visarjanvyasya sah, and so rule 1.1.52 would correctly direct the substitution to the final sound of vrksah, namely to h. But this, Patanjali argues, cannot be called a purpose. The replacement of only the visarjaniya is already taken care of by the nirdisyamdna-paribhdsd, the visarjaniya being expressly enunciated in rule 8.3.34.116 Patanjali thus explicitly says that it is untenable to suggest that rule 1.1.52 alo 'ntyasya is a purpose for reading the ayogavdha sounds in the Sivasutras, for the nirdisyamdna-paribhdsd provides for the correct application of the operations in question. At this stage of his exposition Nagesa leaves objections aside, turning instead to the functions served by the, paribhdsd and its relationship with other rules. He begins by pointing to an issue brought up also in the introductory part
114

115 116

The ayogavdha sounds are visarjaniya (visarga), jihvdmuliya (the sound replacing visarg before k and kh), upadhmdniya (the sound replacing visarga before p and ph), anusvdra (nasal), anundsika (nasalised sound) and yama (doubled consonant before a nasal consonant). A pratydhdra is a technical abbreviation term which serves to denote a set of sounds. For example, iK denotes long or short i u r I. Mbh 1:29,21-4: alo 'ntyavidhihprayojanaml vrksas tarati I plaksas taratil alo 'ntyasya vidhayo bhava alo 'ntyasya satvam siddham bhavatil etad api ndsti prayojanaml nirdisyamdnasyd bhavantiti visarjaniyasyaiva bhavisyatill [A 1.1.52] alo 'ntyasya is a purpose [for teaching the ayogavdha sounds in the Sivasutras so that they can be referred to by the term al]. For example, [s replaces the visarjaniya (h) of vrksah and plaksah according to A 8.3.34 visarjaniyasya sah] in vrksas tarati and plaksas tarati. Since [substitution] operations [by rule 1.1.52] apply to the final al, it is correctly established that s is [the replacement] for the final al. - Even this is no purpose. [In the operation taught by rule 8.3.34, s] will apply only for the visarjaniya by virtue of [the paribhdsd] 'substitutes occur in the place of what is expressly enunciated'.

224

Indian semantic analysis

of his discussion, namely that the paribhdsd prevents units identified by the tadantavidhi and the yaddgama-pnnciple from being recognised as elements which are to undergo a substitution operation. The problems caused by the yaddgama-principlo have already been mentioned. The tadantavidhi is rule 1.1.72 yena vidhis tadantasya which states that when an operation is taught by means of some element, that element denotes, apart from itself, also that which ends with it. For example, the element pad would denote also supdd. To illustrate his point, Nagesa adduces as examples the formation of udasthdt and supadah. The problems related to the correct application of rules in deriving the aorist form udasthdt have been discussed above. The other example, supadah, refers to the operation taught by rule 6.4.130 pddah pat in the context of which Katyayana and Patanjali discuss the paribhdsd. According to this rule, the element pad which is final in ajpresaffbcal base termed bhanl is replaced by pad. Thus, in the masculine, the accusative plural of the presuffixal base supdd 'having beautiful feet' is formed by adding the suffix Sas (-as). The base would then fulfil the conditions for being termed bha, and pad is accordingly replaced by pad, giving at this stage the form supadas. By the tadantavidhi, however, the unit pad would also denote units such as supdd and dvipdd, yielding in the above example just padas and not supadas in the accusative plural. But on the condition that a valid substituend, sthdnin, has to be expressly enunciated, such a problem does not arise. So we need the nirdisyamdna-paribhdsd to give us the knowledge of the lack of sthdnitva, the property of being a substituend, for elements which are suggested solely by the tadantavidhi or the yaddgama-principle. This is recognised already by Katyayana when he introduces the paribhdsd in his third vdrttika on A 6.4.130.118 Once the paribhdsd is accepted, forms such as supadah and udasthdt are properly accomplished. Having thus established the paribhdsd in relation to the tadantavidhi and the yaddgama-pnnciple, Nagesa goes on to remark on its relation to th£ three rules A 1.1.52 alo 'ntyasya, A 1.1.54 ddehparasya, and A 1.1.55 anekdlsit sarvasya. He claims that when each and every condition for thelfpplication of these rules is fulfilled, then they co-exist jointly with the paribhdsd. There is no relation of blocked and blocker since there is no contradiction involved here. Moreover, says Nagesa, they are not subservient to each other, since both th6 paribhdsd and these rules are introduced in grammar for the sake of other rules.119
117

119

According to A 1.4 18 yaci bham, a unit gets the name bha when it is followed by a suffix beginning with a vowel or y, provided the suffix in question fulfils the conditions required by A 1 4.17 svddisv asarvandmasthdne, namely that it should be one of the suffixes introduced by rules A 4 1.2 through A 5.4.151 with the exception of suffixes that are sarvandmasthdna. The latter are the suffixes of the so-called strong cases. In Pamnian terms they are (A 1.1.42) Si, that is, the replacement of Jas and to1 (the nominative and accusative plural suffixes) after a neuter presuffixal base (A 7.1 20), and (A 1.1.43) suT, that is, su, au, Jas, am, auT- the suffixes of the nominative singular, dual, plural,118 accusative singular and dual when they and occur after a non-neuter presuffixal base. See p 218, note 102 above. , See p. 209, note 76 above.

Substitution

225

Now, one may object that there is a possible conflict here with rule 1.1.55 anekdlsit sarvasya. This rule teaches that a substitute marked with a diacritic S or one consisting of more than one sound replaces the entire element which is to undergo substitution, instead of only its final sound as stipulated by rule 1.1.52 alo 'ntyasya. In the formation of supadah, then, the element pad ought to replace the entire base supdd. Nagesa refutes this by claiming that the word sarvasya 'of the entire' in rule A 1.1.55 should be understood so as to denote all of that which is identified as a sthdnin through the nirdisyamdna-paribhdsd only, but not all of that which is suggested by the tadantavidhi or the yaddgama-principle. Moreover, Nagesa continues, to suggest that A 1.1.52 alo 'ntyasya and A 1.1.54 ddeh parasya should block the paribhdsd by reason of their being taught later than A 1.1.49 in the AstddhydyT, is contradicted by a statement in the Mahdbhdsya. Under vdrttika 7 on A 6AA30pddahpat Patafijali comes up with the paribhdsd to solve the difficulties connected with the formation of udasthdt mentioned earlier. A 8.4.61 udah sthdstambhoh purvasya states that after ud, in the place of the initial sound of sthd and stambh, we have a sound which is homogeneous with the earlier. That the substitution takes effect on the initial sound is provided for by rule 1.1.54 ddeh parasya which teaches that when an operation is taught for a following form, such as sthd and stambh which follow ud, then the substitution is directed to the initial sound of that form. Rule 1.1.54 teaches an exception to the stipulation of rule 1.1.52 alo 'ntyasya, that the substitution is directed to the final sound. Since rule 1.1.54 and, accordingly, rule 1.1.52 are involved in the process here, Patafijali would not have brought up the paribhdsd if these rules were to block it. Nagesa is thus relying on the authority of Patanjali in this case. With regard to the opposite idea, that the paribhdsd should block rules A 1.1.52 and A 1.1.54 by force of its being antaranga, Nagesa refers to Kaiyata's statement at A 6.4.142 ti vimsater diti that those two rules would have no scope if this were true.120 Hence, the paribhdsd applies simultaneously with rules A 1.1.52, A 1.1.54 and A 1.1.55. The paribhdsd limits a substitution operation to take effect only on that element which in a rule is expressly enunciated as ending in the genitive case suffix, while the rules mentioned direct the substitution to a specific part of that element. The precise relationship of the paribhdsd to these seemingly conflicting rules is thus clarified. Nagesa next considers a circumstance which precludes the application of the paribhdsd. This concerns the suffix akAC (-ak- with stem-final accent). According to A 5.3.71 avyayasarvandmndm akacprdkteh this suffix is added
120

Kaiyata states (MbhP V354a)# na hi nirdisyamdnapanbhdsd 'alo 'ntyasya' 'ddehparasya' ity etayor bddhikd, etayor mrvisayatvaprasangdt, 'for the nirdisyamdna-paribhdsd do block the two [rules] alo 'ntyasya and ddeh parasya since these two then by consequence would be without any scope' Whether Nagesa agrees with Kaiyata in this respect is hard to tell; cf Kielhorn 1871-4 73, note 5

226

Indian semantic analysis

to an indeclinable word and to a pronoun before the vowel of the final syllable. Thus, for example, the pronominal stem yusmakad- is derived by adding akAC to yusmad-. Defined as a pratyaya 'suffix', and not as an dgama 'augment', akAC is one of the two 'suffixes' which are added inside a form.121 In the domain of akAC, Nagesa claims, the paribhdsd does not apply. The reason is that a substitution operation takes effect only when we have a knowledge of the sthdnin involved, as does sthdnivadbhdva, the principle whereby a substitute is treated like the element it replaces. Similarly, in the case of akAC we get a knowledge of the sthdnin by another paribhdsd, presented as paribhdsd 89 in the Paribhdsendusekhara: tanmadhyapatitas tadgrahanena grhyate. This paribhdsd states that something which is inserted in the middle of a form is understood when this form is mentioned. This applies to akAC, and a substitution operation takes effect only when we have a knowledge\tf the form to which akAC is added. Thus, what is taught for yusmad- apples also to yusmakad-. As long as akAC is inserted in a form the nirdisyamdna-paribhdsd does not apply. The form to which akAC is added is recognised as the sthdnin, the element which is to undergo some future substitution operation, by paribhdsd 89.122 The final issue brought up by Nagesa concerns cases where a genitive form has been expressly enunciated in a rule where it thus exhibits a sthdnasasthT and then recurs by anuvrtti in a subsequent rule where it has to be, interpreted as exhibiting an avayavasasthi, a partitive genitive. Nagesa claims that the nirdisyamdna-paribhdsd does apply even in the domain of such partitive genitives. The examples Nagesa cites are the derivations of atisyah^ nominative singular masculine of the stem atityad- 'surpassing that', and atiyuyam, nominative plural of the stem atiyusmad- 'surpassing you'. As for atisyah, A 7.2.106 tadoh sah sdv anantayoh states that t and d which are non-final in a presuffixal base are replaced by s in the nominative singular. More specifically, this applies to the non-final t and d of the pronominal stems tyad- etc. as stipulated by A 7.2.102 tyadddindm ah which teaches that a replaces the final (A 1.1.52) of the presuffixal bases (A 6.4.1) tyad etc.123 In rule 7.2.102 the genitive tyadddindm is to be interpreted as a sthdnasasthT, whereas it is to be interpreted as an avayavasasthi, a partitive genitive, in the context of rule 7.2.106. Now, in the formation of the nominative atisyah from the stem atityad-, the substitute s prescribed by rule 7.2.106 replaces the t of
121 122 123

The other one is SnaM, the nasal affix of the seventh present class (A 3 1 78 rudhddibhyah snam) For the view of Raghavendracarya, see Kielhorn 1871-4 74, note 2. See also the discussion in the Mahabhasya at varttika 1 on rule A 1.1.72 (Mbh 1.184,13-25). The nominative of tyad- is derived as follows : tyad+sU (A 412) tya + a + sU (A 7.2.102) sya + a + sU (A 7 2 106) sya + sU (A 6 197) syas

Substitution

227

tyad only, not the t of ati although tyad would denote also atityad by rule 1.1.72 yena vidhis tadantasya. The reason is that only tyad etc. have been expressly enunciated in tyadddindm, which in the context of rule 7.2.106 is a partitive genitive. It should however be stressed that the paribhdsd applies only with regard to the expressly enunciated genitive tyadddindm of rule 7.2.102, not to any genitive. The second example offered by Nagesa is the formation of atiyuyam. We are here dealing with rules that teach substitution operations in the domain of pronouns. According to A 7.2.86 yusmadasmador anddese, the finals of the two presuffixal bases yusmad and asmad are replaced by a before a case suffix which begins with a consonant and is not a substitute. A 7.2.91 maparyantasya states that the substitutes taught in the subsequent rules up to A 7.2.98 replace that part of yusmad and asmad which is bounded by mf that is to say, yusm and asm. By A 7.2.93 yuyavayau jasi, then, yuya and vaya are taught as substitutes for yusm and asm in the nominative plural, yuyam and vayam. Now, the compound yusmadasmadoh exhibits a sthdnasasthiin A 7.2.86, but when it recurs by anuvrtti in rules A 7.2.91 and A 7.2.93 it has to be interpreted as a partitive genitive. If the paribhdsd did not apply in the case of such a partitive genitive, then, in the formation of atiyuyam, the unit yuya would replace even atiyusm by rules A 7.2.91 and A 7.2.93 since here yusmad would denote also atiyusmad according to the tadantavidhi, rule A 1.1.72 yena vidhis tadantasya. On this evidence Nagesa concludes that the nirdisyamdna-paribhdsd is to be considered applicable also in cases where substitutes replace only part of that which is expressly enunciated in a rule with a genitive case ending, that is, the paribhdsd applies also to what is expressly enunciated when this recurs by anuvrtti in a subsequent rule even though the genitive in the latter context is to be interpreted as a partitive genitive. Nagesa ends his exposition by pointing out that the paribhdsd is found in the Mahdbhdsya on A 6.4.130 pddah pat and A 1.1.49 sasthT sthdneyogd. The major points of Nagesa's discussion may be reviewed briefly as follows. He introduces the paribhdsd by referring to a difficulty caused by the yaddgama-prmciple presented in the preceding paribhdsd 11, namely that a linguistic unit recognised according to this principle may erroneously be identified as a unit which is to undergo a substitution operation. This problem is solved if the nirdisyamdna-paribhdsd is established. Nagesa goes on to claim that rule 1.1.49 is read twice and that the paribhdsd is established by the second reading. Having thus tentatively established the paribhdsd, Nagesa states various objections against it which have to be met. The first of these concerns substitution operations which take effect even on long vowels while only the corresponding short vowel has been expressly enunciated in the rule which teaches the operation. The second objection concerns the possible redundancy of the anubandha N which, among other functions, serves to signal that a substitution operation is to take effect only on the final sound of the element identified

228

Indian semantic analysis

as the substituend. Both of these objections are successfully met by Nagesa who, significantly, relies on the authority of Patanjah in this endeavour. At this stage Nagesa goes on to remark on the relationship between the paribhdsd and certain other rules of grammar with which there may be a possible conflict. His first point here is that the paribhdsd is necessary to prevent elements suggested solely by A 1.1.72 yena vidhis tadaniasya or the yaddgama-prmciple from being recognised as elements which are to undergo substitution operations. Nagesa's second point here concerns the precise relation between the paribhdsd and the three rules A 1.1.52 alo 'ntyasya, A 1.1.54 ddehparasya and A 1.1.55 anekdlsit sarvasya. He argues that neither do these rules and the paribhdsd block each other, nor are they subservient to each other. Provided each and every condition for their application is fulfilled, the rules and the paribhdsd apply together. The paribhdsd limits a substitution to take effect only on that which is expressly enunciated with a genitive pase ending, while the rules mentioned direct the substitution to a specific part of that which is expressly enunciated. Nagesa then points out an exception to the applicability of the paribhdsd, namely forms to which the suffix akAC has been added. The property of being a substituend, sthdnitva, is ascribed to such forms by another paribhdsd, presented as paribhdsd 89 in the Paribhdsendusekhara: tanmadhyapatitas tadgrahanend grhyate. The final issue raised by Nagesa concerns the applicability ofthe paribhdsd in the domain of partitive genitives. Given a rule where a form is expressly enunciated with a genitive case ending which in that context is to be interpreted as a sthdnasasthi, the same form may recur by anuvrtti in a subsequent rule where it has to be considered a partitive genitive. Nagesa argues that the paribhdsd does apply in the domain of such partitive genitives, that is, it applies even in cases where the substitute replaces only part of that which is ? expressly enunciated in a rule. In this way the interpretation which Nagesa offered at the beginning of his discussion has been accounted for. The passage in question runs as follows (ed. Abhyankar 1962:28,11-12): sasthyantam nirdisyamdnam uccdryamdnam uccaryamanasajatiyam $va nirdisyamdndvayavarupam eva vd sthdnena sthdnamrupitasambandhenq yujyate na pratiyamdnam iti tenedam siddhamll What is expressly enunciated - that is, actually pronounced, [or rather, what m the real language] is of the same kind as, that which is actually pronounced - with a genitive case ending, only [that] or only that which forms part of what is expressly enunciated, is linked with124 the sthdna, that is, through the relation which i£ determined by the sthdna, not that which is [merely] understood [by the tadantavidhi or the yadagama-ynncvplz]. Hence this [rule A 1.1.49] is established.
124

The sense of the instrumental here is ambiguous, but the whole issue of the relation defined by the sthdna will be discussed below

Substitution

229

In this way the panbhdsd does not only convey the sense of rule 1.1.49. It also serves to reconcile rule 1.1.49 thus interpreted with certain other features of the system set up m the AstddhydyT. It may be noted in passing that Nagesa considers the two terms of the ^/zana-relation to be related through the sthdna in the sense that the relation is determined by the sthdna or, as will become evident later on, by what the word sthdna signifies. This issue as well as the conclusions that may be drawn from Nagesa's discussion of the nirdisyamdnaparibhdsd will be considered again below. Commentators' remarks on the final section of the Mahdbhdsya at A 1.1.49 and their implications With this background it is possible to turn to the concluding remarks of the commentators on the final passage of the Bhdsya at A 1.1.49. Patanjali, it may be recalled, stated that the purpose of rule 1.1.49 is to render the nirdisyamdnaparibhdsd.125 Bhartrhari remarks as follows in the DTpikd:126 sasthyantam sthanena yatha yujyeta/ yat tat sasthyantam srutyanupdti I yat pratiyamdnam tan nam sthanena yujyeta ml yatah sasthy uccarita/ tasya kdryayogitvamm ity eso 'rthah sutrdrambhdt labhyate itil lyam ca sd panbhdsd 'nirdisyamdnasyddesd bhavanti' itil tena 'pddahpaf iti pdcchabdah sasthyantah srutyanupdti, tadantah pratlyamdna iti na tadantah sthanena yujyatel nanu ca pdda ity etat sambhutam129 tadantam pratydyayatiti tadantasyddesaprasangahl tatra kecid varnayantil tadantapanbhdsd bddhyate 'nenetil etac cdsatl iha dosah sydtl pddayater apratyaydntasya padbhdvah sydt130/ iha hi tat sydn na vydghrapdda itim I tasmdt prdptasya132 tadantavidher anena vyavacchedah knyatel yathaiva sdstrdt pdda iti pddantam pratipadyase, evam asmdc chdstrdd133 idam
125 126

Seep 218 above Text established on the basis of MbhDms fol 204b-5a, MbhD 293,10-23, and the improved edition of Palsule and Bhagavat 1991 23,18-24,5 127 So read with MbhDms and Palsule and Bhagavat, against the reading ° srutyanupdti yat pratiyamdnam tat taf of MbhD Why Abhyankar and Limaye introduced this emendation remains obscure in as much as one would have to translate 'whatever reaches the ear with a genitive case ending and is so understood, that should be joined with the sthdna'' This goes against the very idea of preventing what is understood by the tadantavidhi or the yaddgama principle from being recognised as elements which are to undergo a substitution operation It would probably be possible to regard the statement as too wide an application (atiprasanga) which is later to be rejected, though the MbhDms is completely clear at this point and makes perfect sense, so there is no need to emend here 128 So read with Palsule and Bhagavat against kdryayogitve of MbhD and MbhDms 129 Palsule and Bhagavat suggest the emendation samsrutam0, but this does not make very good 13 sense either ° MbhDms padbhdva syd 131 The MbhDms seems to read iha hi tasyd iha hi tasydnna vydghrapdda iti This is so garbled ;„ that it is tempting to follow Palsule and Bhagavat who simply suggest iha hi na sydt '.j vydghrapdd iti 132 Emendation suggested by Palsule and Bhagavat, MbhDms reads tasmdn prdtastha, MbhD emends to tasmdd pddasya 133 MbhDms seems to read evam ayam vdcchdstrdd, Palsule and Bhagavat suggest evam eva vdsmdc chdstrdd (reading, however, cdsmdc°, ibid 195), MbhD reads evam ayam vd sdstrdd which is closer to the manuscript reading but makes no sense

230

Indian semantic analysis

^

vdkyam evambhutam pratipadyasva I pdda iti pddantasya yah pdcchabdas134 tasya padbhdvo bhavatltil atha vd samuddyd135 eva kdryasattve pdcchdstrdd136 avayavadvdrakam kdryabhdktvam pratipadyante na sdksdd iti I katham 'yivarnayor didhivevyoK iti I tatrdpi didhivevyor nirdisyamdnatvdt sarvddeso Ippah prdpnotil satyam evam etadl upasamhdrena ca 'alo 'ntyasycC ity antyasya131 bhavisyatill 'In order that what ends in a genitive case suffix should be linked with the sthdnd*: that is to say, whatever ends in a genitive case suffix as a result of hearing [the genitive suffix]; but what is [merely] understood [from that which is actually heard with a genitive case ending], that should not be linked with the sthdna. 'Provided a genitive is actually pronounced after it': that is the recipient of the operation. This sense is obtained from the formulation of the rule [A 1.1.49]. And this is [as much as ] that [familiar] paribhdsd: 'substitutes occur in the place of what is expressly enunciated'. Therefore, in [A 6.4.130] pddah pat it is the linguistic item pad which ends in a genitive case suffix as a result of hearing [it]; what ends in it (= in pad) [and would thus fall under the scope of the tadantavidhi (A 1.1.72)] is [merely] understood, and hence that which ends in it is ncjt linked with the sthdna. Well, but this [element] pddah has come about, [and that] causes a knowledge [also] of what ends in it [according to A 1.1.72], and so there is a possibility that the substitute occurs in the place of whatever ends in it (= in pad). » As far as this is concerned, some explain: the tadanta-paribhdsd is blocked by this [rule 1.1.49]. But that is not correct; there would be a fault here: pad would occur in the place of [the causative presuffixal base] pddayatih {-pad) ending in a zero suffix {-pad). But there would not be [a fault] here: [pad would not occur] in the place of vydghrapdd. Therefore this [rule 1.1.49] accomplishes a delimitation of the tadantavidhi which applies [here]. Just as, on the basis of the rule (=A 1.1.72), you understand [also] whatever ends in pad when you hear [the word] pddah [of A 6.4.130], in the same way, on the basis of this rule (= A 1.1.49), you must understand this sentence [pddah pat] in this manner: pad occurs in the place 6f the linguistic element pad of that which ends in pad [when you consider] pddah'in the place of pad' [in rule 6.4.130]. Or rather, on the basis of the pad-rule [i.eh, A 6.4.130138], when a grammatical operation takes place, entire groups of [linguistic] elements themselves are capable of participating in a grammatical operation, [but only] through [their] parts, [and] not directly. ?, So read with Palsule and Bhagavat; MbhDms seems to read pddacchabdas°, emended to pad sabdas MbhD. MbhDms reads yamuyd; Palsule and Bhagavat accept the emendation samuddyd[h] 'wholes' of MbhD. The emendation is substantiated by the subsequent avayavadvdmkam 'occasioned E by the parts'. 136 With MbhD. MbhDms seems to read yd[c?]chdsdstrdd (according to Mbht), pdchasdstrdd; according to Palsule and Bhagavat, ydchdstrdd). Palsule and Bhagavat suggest the emendation {kdryasattve) 'smdc chdstrdd; see below. 137 After upasamhdrena MbhDms seems to read valomtyasyetyamsya; MbhD gives the text as upasamhdrena ca 'alo 'ntyasyd1 ity antasya; Palsule and Bhagavat as above* except that they read tu for ca. \ 138 But it would probably be neater to have A 1.1.49 as the basis here. This, I presume, is what makes Palsule and Bhagavat prefer the more drastic emendation {kdryasattve) 'smdc chdstrdd0. *'
135 134

Substitution

231

What about [A 7.4.53] yivarnayor didhfvevyohl139 There too, because didhi and vevf have been expressly enunciated, elision applies as the substitute of the entire [forms didhTand vevT]. That is correct, but by invoking [A 1.1.52] alo 'ntyasya, [the substitute] will appear [only] in the place of the final sound. Once again, the task is to get the relationship right between A 1.1.49, or the nirdisyarnana-paribhdsd, and certain other rules in the AstddhydyT. Bhartrhari limits himself to the seemingly conflicting tadantavidhi and rule 1.1.52 alo 'ntyasya. Because didhi and vevT have been expressly enunciated - as stipulated in the paribhdsd - A 1.1.52 directs the substitution to the final sound, and the tadantavidhi which would bring about substitution of the entire forms is set aside. Kaiyata, who is considerably clearer here, comes up with a slightly different solution while moving within the same scenery (MbhP I:41?2b): yatah sasthlft'/ tena pddah pad ity asydyam arthah -pddantasydngasya yo 'vayavah pdcchabdah sutre sasthyd nirdistas tasya pacchabda ddeso bhavati, na tu pratiyamdnasya pddantasyal asyaiva sutrasya nirdisyamdnasyddesd bhavantlty ayam arthah I sasthiti sasthyantam grhyate I tena sasthyantam eva nirdisyamdnam sthdnena yujyate, na tu pratlyamdnam ity arthah I alo 'ntyasyddeh parasyety etat tu vacanasdmarthydn na bddhyate II That after which the genitive case suffix [is actually pronounced]'; hence, the meaning of [rule 6.4.130] pddah pat is the following. It is as a replacement of that particular unit pad, which forms part of a presuffixal base ending in pad and is expressly enunciated in the rule by the genitive case suffix, that the unit pad applies, but not of that which is [merely] understood to end in pad. The purpose of this very rule [A 1.1.49] is the statement that 'substitutes occur in the place of what is expressly enunciated'. In saying 'a genitive case suffix', what is denoted is 'that which ends in a genitive case suffix'. Therefore, only when it ends in a genitive case suffix is something expressly enunciated linked with the sthdna, not when it is [merely] understood [by the tadantavidhi or the yaddgama-pnnciple]; that is the meaning. But [rules A 1.1.52] alo 'ntyasya and [A 1.1.54] ddeh parasya, by force of [the very fact that they] have been stated, are not blocked140 [by A 1.1.49]. Adducing A 6.4.130 pddah pat as his example, Kaiyata states that only when something is expressly enunciated with a genitive case ending is it recognised as a valid sthdnin, not when it is merely understood to be so. He also notes that when rule 1.1.49 is interpreted in this way it does not block rules A 1.1.52 alo 'ntyasya and A 1.1.54 ddehparasya. By now this is familiar ground. Nagesa's commentary contains remarks of considerable importance here (MbhU I:412b):
139

A 7.4.53 teaches elision in the form of lopa as the replacement of the final sound (A 1.1.52 alo 'ntyasya) of the roots didhiN and veviN when they occur as presuffixal bases before suffixes beginning with y or i. Under the solution presently discussed, lopa would be the substitute of the entire forms didhi and vevu I4° See p. 225, note 120 above.

232

Indian semantic analysis

nirdisyamdnam itil uccdryamdnam ity arthahl nirdisyamdnam sasthyantam evety anvayahl nirdisyamdnapadasya cddhydhdra iti bhdvahll na bddhyate itil kim tu svavyavasthdpakatvendsnyata ity arthahl evam anekdlsutre sarvo 'py etatparibhdsdbodhita eva grhyate I paddanna itydddv api ddesaih pdninikrtasthdnyuccdrandnumdndd vyavasthdrthatvdd vd na dosah II yat tv alo 'ntyasyddeh parasyety amu etadbddhakdv iti, tat tu udasthd iti141 pddahpad iti sutrasthabhdsyaviruddham I vyavasthdpakatve sambhavati bddhakatvakalpand'naucityagrastamlA^call 'What is expressly enunciated'; that is to say, what is actually pronounced. The construction of the sentence [in the Pradipa] is: 'only what is expressly enunciated with a genitive case ending',143 the idea being that we have the supplement of the word nirdisyamdna 'being expressly enunciated'. '[Rules A 1.1.52 and A 1.1.54] are not blocked', however these [two rules] are relied upon as limitors of the paribhdsd,144 that is the sense. Likewise in rule [A 1.1.55] anekdl[sit sarvasyai 'the entire' [unit identified as the sthdnin by this rule] is also understood to be only/what is recognised by this paribhdsd.145 Even in the case of [A 6.1.63] paddannqs etc. there is no fault, either because the pronunciation of the substituends actually/ made by Panini is inferred through the substitutes, or because the purpose [of the fule] is a fixed distribution [of usage]. But the proposition that the two [rules A 1.1.52] alo 'ntyasya and [A 1.1.54] ddeh parasya block the paribhdsd is contradicted by the Bhdsya at rule [A 6.4.130] pddah pat with regard to [the formation of an aorist form such as] udasthdtdm.146 Moreover, this [proposition] is tormented due to the impropriety in imagining that [these rules] serve as blockers [of the paribhdsd] when the possibility exists that [they] serve to determine a fixed distribution of scope. Nagesa here points out that even in the case of A 6.1.63 paddannomdshrnnisasanyusandosanyakanchakanudandsanchasprabhrtisu there is no fault. This rule teaches that before one of the suffixes beginning with Sas, that is, before the accusative plural suffix and the other suffixes of the so-called weak cases, the substitutes listed should replace the stem, so that pad replaces pdda, dat replaces danta, etc. This rule is an anomaly in that substitutes are taught, but substituends are not listed. And when no sthdnin has been expressly enun141 142 143

144

145

146

The edition of Vedavrata (1962-3) reads udasthdd iti.' ~ ~ The edition of Vedavrata (1962-3) reads bddhakatvakalpandnaucityagrastam which I take to be a compound. Kaiyata states 'sasthyantam eva nirdisyamdnam'. This syntax is determined by the logic of Kaiyata's explanation. Nagesa, on the other hand, prefers the syntax 'nirdisyamdnam sasthyantam eva\ I suspect that Nagesa's concern here is to make it clear that the particle eva in the sense of restriction, avadhdrana, should apply to both of the other words of the construction. This is confirmed by his own word order m the Paribhdsendusekhara (ed. Abhyankar 1962:28,11); sasthyantam nirdisyamdnam ... eva. sva- may refer to rule 1.1.49 itself, which at this stage of the discussion amounts to the same. It is the scope of rule 1.1.49, interpreted as conveying the sense of the paribhdsd, that is limited by the two rules; that is, the two rules determine a fixed distribution of scope. Just as rules A 11 52 and A 1.1.54 serve to direct the substitution to a specific part of the unit which has been identified by the paribhdsd, so does rule 1.1.55 state that the substitute under certain conditions is to replace the entire substituend, but only what has been identified by the paribhdsd, not that which has been recognised by the tadantavidhi or the yaddgamaprmciple. This variant of the example is the one adduced by Patanjah in his discussion of A 6.4.130 (Mbh 111:222,12).

Substitution

233

ciated with a genitive case ending, the paribhasa does not apply. Nagesa gets around this problem in two ways. Either, he says, the pronunciation of the substituends made by Panini can be inferred through the substitutes, or the rule serves to establish a fixed distribution or limitation (yyavastha) in usage. Similarly, when dealing with the possibility that the rules A 1.1.52 alo 'ntyasya etc. were to block the paribhasa, he concludes that such a view is untenable as long as the possibility exists that these rules serve as limitors of the paribhasa, determining a fixed distribution of scope. Vaidyanatha Payagunda comments on the Uddyota passage in the following way (Bhargavasastrl JosT 1951:412, note 10): nanv ayam paddanna ityddau dosah, atra sthdnyabhdvenaitadapravrttyd tesdm sarvapadendgrahanendntyddesatvdpatter ata aha — paddanna itydddv itill atra gauravam, tathd sati tathaiva vaded vd yta aha - vyavetil prayogetyddihll dvayor api loke sattvdd iti bhdvahll Well, there is this fault in the case of [A 6.1.63] paddannas etc.: since no elements which are to undergo substitution are mentioned here, the paribhasa does not apply. Hence the word sarva [of A 1.1.55] is not denotative, and so you would have the substitution directed to the finals of these [by rule 1.1.52]. Therefore [Nagesa] says 'in the case of [rule 6.1.63] paddannas etc.'. There is prolixity involved here, or, that being so, that is indeed how he would say [it], [and] therefore he states: '[or because the purpose is] a fixed distribution'; that is, in usage etc. The idea is: because both exist in the world. Units occur in complementary distribution. Substituends would already exist in the world, but we use substitutes before certain case suffixes. Rule 6.1.63 serves to establish a fixed distribution in this respect. This is possible because sthdnins and ddesas both exist in the world, a fact which implies that one has to consider usage. And Paninlyas do indeed acknowledge that grammar serves to explain the linguistic elements of eternal Sanskrit usage. Concluding remarks on the discussion of A 1.1.49 Since a survey of Patanjali's remarks on A 1.1.49 was given prior to their discussion, I shall turn directly to the conclusions I think we can draw from this discussion. What is immediately striking is that the series of arguments concerning this rule do not primarily concern the genitive being used to mark the relation of being in the place of. Such a usage of the genitive is in accordance with established Sanskrit usage. Another important conclusion is really a simple observation: in the grammar one and the same genitive form can be interpreted in different ways according to context. In the words of Cardona (1974a:326): Tanini's rules 1.1.49, 66, 67 are properly interpreted only as restrictive metarules. As such they presuppose no radical departure in Panini's metalanguage from the syntax of normal Sanskrit for which the rules of the grammar account.'

234

Indian semantic analysis

Patanjali concludes that rule 1.1.49 is necessary only to teach the content of the nirdisyamdna-paribhdsd. He argues in a similar manner also with regard to A 1.1.72 yena vidhis tadantasya (the tadantavidhi) and A 1.4.13 yasmdt pratyayavidhis tadddipratyaye 'ngam.141 Patanjali is forced to admit that there is some doubt as to the relation marked by the genitive, since between one linguistic unit and another such relations as dnantarya 'adjacency' and sdmipya 'nearness' may also obtain. This he dismisses as mere doubt, and invokes the paribhdsd which tells us that in a case of doubt, we have to ascertain a particular conclusion by reasoned explanation, vydkhydna. So we are forced to interpret and in cases of doubt one can rely on tradition. Rhetorically, then, he says we do not need rule 1.1.49. Then he says we do need it, but only to render the nirdisyamdna-paribhdsd. In other words, the rule teaches specific needs of grammar and not a particular usage of the genitive. This argument wou|d not work were not the use of the sthdnasasthT firmly rooted in the usage of ordinary Sanskrit. It seems therefore safe to conclude that the usage of the sthdnasasthT is an established feature of ordinary language and that Patanjali's discussion of rule 1.1.49 supports this fact. This leads to the question of whether Patanjali's conclusion is in agreement with Panini's intentions or whether it is a trick of the commentators. That it was Panini's intention that rule 1.1.49 teach what the paribhdsd states is more than doubtful. It is evident that you could get the usage of the sthdnasasthTfrom the Sanskrit language itself, but other interpretations of the genitive also apply in the AstddhydyT, so it is perfectly possible that Panini formulated the rule to clarify this situation. Otherwise one would have to assume that he introduced rule 1.1.49 to clear the ground for rules A 1.1.52, 1.1.54, 1.1.55 and 1.1.56 which all pertain to substitution. Commentators suggest that A 1.1.56 is ajfidpaka for A 1.1.49, but so would A 1.1.52 alo 'ntyaysa be. It may not technically be zjiidpaka, but could it be that rule 1.1.49 w&s stated to account for the genitive in that rule and the ones in A 1.1.54 and A 1.1.55 which teach exceptions to rule 1.1.52? Katyayana gives no hint that the purpose of A 1.1.49 is to teach the content of the paribhdsd. On the contrary, he is concerned with the genitive and its usage in the AstddhydyT. The four vdrttikas on this rule state: (1) that the purpose is to make a restriction with regard to the interpretation of a genitive in the grammar (2) that this does not work, since it would exclude such genitives as the partitive (3) that this is no problem, because in those cases there is no doubt about the relation (4) or, that the genitive used to denote the relation defined by a sthdnasasthT should receive some distinguishing mark. The Paninlya tradition from Patanjali to Vaidyanatha Payagunda, on the other hand, tells us that A 1.1.49 is introduced to convey the sense of the nirdisyamdna-paribhdsd, that substitutes take the place of what is expressly enunciated with a genitive case ending. Simple commentaries, such as the
147

For the latter rule, see Benson 1990.70-1.

Substitution

235

Kdsikdvrtti and the Bhdsdvrtti, do not bring in the paribhdsd in their discussions of rule 1.1.49. The need of grammar, however, is that in the process of a grammatical derivation a rule such as A 6.4.130 pddah pat teaches that pat replaces pad but not anything ending in pad which has not been stated with a genitive case ending. This may be achieved by letting A 1.1.49 render the paribhdsd, but that could only happen if rule 1.1.49 were dispensable in its function of introducing the sthdnasasthT. In his long discussion oftheparibhdsd in the Paribhdsendusekhara, Nagesa is not really concerned with substitution, only with the technical complications caused by the paribhdsd in the system of grammar. Consequently, he discusses problems internal to the AstddhydyT, primarily the relation of the paribhdsd to other rules, in particular rules teaching or relating to substitution operations. The paribhdsd limits a substitution operation to take effect only on that element which in a rule is expressly enunciated as ending in a genitive case suffix, while the rules in question direct the substitution to a specific part of that element. The precise relationship of the paribhdsd to seemingly conflicting rules is thus clarified. This raises the important question of whether the paribhdsd is necessary for non-grammarians. Although this is hard to argue, I am inclined to answer maybe not, and say that it is only required within the theoretical framework of Paninlya grammar. Strictly speaking the paribhdsd speaks of 'what has been enunciated', but not of the genitive. It is the choice of the grammarians to let it render A 1.1.49. If the paribhdsd has any general validity, it does not necessarily state anything more than that a substitute takes the place of an expressly enunciated substituend. As for the sthdnasambandha, the relation defined by supplying the locative word form sthdne, Kaiyata tells us that it is antaranga, that it is what would come to mind anyway as the relation obtaining between one linguistic element and another regardless of other relations that may possibly obtain. This view is endorsed by Bhattoji DIksita, Nagesa, and Vaidyanatha. Nagesa comments that antarangatva is established regardless of whether we interpret the term sthdna as artha 'meaning' or prasanga 'possible appearance', the two meanings most commonly ascribed to this term by the tradition. If it signifies artha 'meaning', Vaidyanatha remarks, then the antarangatva of the sthdnasambandha holds because meaning comes first when we use words. But if it is assumed that sthdna signifies artha, then one does not need Panini's rule, since words are used to denote or express artha and that is in the mind already. If, on the other hand, sthdna signifies prasanga 'possible appearance' it is also possible to interpret the genitive as a sthdnasasthl'even without rule 1.1.49. The possibility of other relations, Vaidyanatha argues, exists only for what is prasakta, for what would possibly appear. Other relations, such as adjacency and nearness, can obtain only when something appears. These relations hold only between word and word, not between word and meaning. But either way antarangatva is established. The alternative senses of artha and prasanga are

236

Indian semantic analysis

just choices with regard to the weight being laid on meaning or on word form, though prasanga elsewhere involves meaning. If, then, the sthanasambandha is considered to be present in the mind before any other relation when we meet with a genitive form in the grammar, there is every reason to conclude that to interpret a genitive as a substitutional genitive is nothing remarkable in Sanskrit grammatical literature. Nor is there anything in the discussions of A 1.1.49 and the nirdisyamdna-paribhdsd which restricts such a usage of the genitive to grammatical texts. With regard to the relation sustained between the two terms of a nirvacana analysis, this is significant. So is Vaidyanatha's remark that on the view that the term sthdna signifies artha 'meaning', the sthanasambandha is what comes to mind before any other relation because meaning comes first when we use words. It may be recalled that nirvacana analysis first and foremost is concerned with meaning, not with grammatical form. It remains to investigate more precisely the nature of the sthanasambandha, the relation marked by a sthdnasastlu. The relation marked by a sthanasasthi The logical possibilities are that there may exist in the substitution model three relations: (a) between the sthdnin and the ddesa (b) between the sthdnin and the sthdna (c) between the sthdna and the ddesa. It may also be the case that sthdna is a relational term, that is, that it is expressive of the relation 'being the sthdna of. In the Mahdbhdsya discussion of A 1.1.49 we meet with some conflicting statements as to the relation marked by a sthanasasthiand its interpretation. To begin with, Patanjali gives two alternative interpretations when he discusses the vyadhikaranabahuvnhi compound sthdneyogd (Mbh 1:118,6-7): sthdne yogo 'sydh, literally 'that of which there is a relation in the sthdna", and sthdnena yogo 'sydh, 'that of which there is a relation with the sthdna9. The locative interpretation I take to mean that the sthdna is a necessary condition for the relation, while the instrumental would imply that the sthdna itself is one of the terms of the relation. This ambiguity reflects itself in Patanjali's final remarks on A 1.1.49: sthdna iti vydkhydsydmah, 'so we shall interpret [the genitive to denote a relation defined by the locative word form] sthdne' (Mbh 1:119,25), and sasthyantam sthdnena yathd yujyeta, 'so that what ends in a genitive case suffix should be linked with the sthdna' (Mbh 1:119,26). In works such as the Praudhamanoramd and the Tattvabodhinlonly the trtiydvigraha is resorted to.148 As I have already suggested,149 this could simply be prompted by the fact that one would expect the word yoga in the sense of sambandha 'relation' to be used with the instrumental form sthdnena in the present context. The most natural way to translate sthdnena yujyeta is 'should be joined with the sthdna9. This is normal Sanskrit usage. Patanjali resorts to it also else148

See p 196 above.

149

See p. 196 with note 50 above.

Substitution

237

where, for example at A 2.2.19 when referring to two views concerning the priority of relations between preverbs, roots, and suffixes (Mbh 1:217,22-3): purvam dhdtur upasargena yujyate pascdt sddhanena, 'a root is first joined with a preverb and subsequently with a finite verb suffix [expressing the kdraka]\ Under this alternative, then, the relation obtains between the sthdna and the sthdnin. There is nothing in the Bhdsya discussion itself which precludes such an interpretation. But the mere fact that the usage of the instrumental is normal Sanskrit usage may allow for the possibility that it is actually resorted to for that very reason, for we did indeed face a locative in the first place. The possibility therefore also presents itself that the expression sthdnena yujyeta is to be interpreted as 'should be joined through/by means of the sthdna\ But there are further conflicting remarks. For example, Kaiyata says at one point in the discussion that the sthdnin is related to the ddesa.150 At another point Nagesa comes up with an ambiguous remark that either the sthdna is one of the terms of the relation in question or it is something which restricts one or both terms of the relation in question.151 A 2.3.50 sasthi sese teaches that a genitive case ending is introduced to denote 'the rest' (sesa). According to the Kdsikdvrtti this rest is any relation, sambandha, which is not a kdraka-relation and different from the meaning of the nominal stem. In other words, a genitive case suffix is introduced to denote any relation sustained between entities, that is to say, any non-verbal relation in general, such as father-son, master-servant, part-whole, etc.152 The fact that two entities are mutually related by their appearance in a given context is expressed by the genitive case. But the particular type of relation is not specified. The problem of determining the relation marked by a sthdnasasthT may best be illustrated by adducing as an example a rule for the interpretation of which rule 1.1.49 comes into play. Such a rule is A 2.4.52 aster bhuh. This rule, together with A 1.1.49, teaches that the verbal root as is replaced by bhu in the domain of drdhadhdtuka suffixes. To the class of drdhadhdtuka suffixes belong, for example, the suffix Ktvd by which the absolutive bhutvd is formed,
150

MbhP I 411a. ddesahngdsangdt tasya ca sasthyabhdvdt tatsambaddhasthdnino lingakd phahsyatity arthah\ 'that is to say, by adding the mark to the substitute, and by the fact that this does not exhibit a genitive case suffix, the operation [signalled] by the mark will take effect on the substituend which is related to the substitute', see p 211 above. 151 MbhU I:410a. yogavatlfr pakse 'py evam eva, sthdnarupasambandhiniyamdrthatvdd asy iU yogavatT"' even on this view things are just the same, since [the rule] has as its purpose a restriction with regard to the relatum having the form of the sthdna\ see p 207 above 152 The Kdsikd states (Kas II 209-11) karmddibhyo yo 'nyah prdtipadikdrthavyatinktah svasvdmisambandhddih sesas tatra sasthi vibhaktir bhavati, 'the rest is such relations owner-owned which is other than [fozrafoz-relations] such as karman "object" [and] different from the meaning of the nominal stem [indicated in the nominative case], in this context the sixth case ending applies'. More sophisticated commentaries, from the Mahdbhdsya onwards, see the need to continue also kdraka-vzlditions from the previous rules, but this need not concern us here.

238

Indian semantic analysis

and the suffix tumUN by which the infinitive bhavitum is formed. In the rule aster bhuh, it is obvious that the statement involves a relation between two relata, the verbal roots as and bhu. Just what the relation is, however, cannot be known from the statement aster bhuh alone, anymore than one knows from the statement devadattasya yajnadattah just what relation obtains between Devadatta and Yajnadatta. But if yajnadattah is replaced by an obvious relational term such zsputrah 'son', the relation in question is immediately understood: Devadatta's son. Now, in the case of aster bhuh, where bhu itself is no obvious relational term, the relation marked by the genitive is determined by the expression sthdne which defines the genitive according to A 1.1.49. A 2.4.52, then, teaches that bhu occurs in the sthdna of as in certain contexts, the elements bhu and as technically termed the ddesa 'substitute' and the sthdnin 'substituend', respectively. But the exact nature of this relation depends on what sense is attributed to the term sthdna. A full investigation of the differenUnterpretations of sthdna considered by the commentators will be carried out sepaJatelyJ^elow. Suffice it here to say that Paninlyas interpret the term sthdna either as an action noun, equivalent to sthiti 'a standing', or as a term referring to a locus. Under the first alternative various contextual meanings are considered, the one most commonly accepted being prasanga 'possible appearance; chance'. Under the second alternative one particular locus that is considered is the artha 'meaning' of a linguistic element or speech unit. If sthdna is interpreted as a term referring to a locus and sthdnin as something occurring in that locus, it seems reasonable to assume that a similar relation obtains between the substitute, ddesa, and the sthdna. The sthdnin and the ddesa are thus inseparably related, but only through the sthdna. Thus, if the sthdna denotes some sort of locus in grammatical space, for example the meaning {artha) in which a linguistic unit is said to occur, then the rules of the AstddhydyT which contain genitives to be interpreted by A 1.1.49 may be assumed to teach linguistic elements that should occur in loci held by other linguistic elements that are to be replaced. If the term sthdna is interpreted as an action noun with the sense of prasanga 'possible appearance' and the term sthdnin is interpreted as 'that which is characterised by possible appearance', then the direction of the relation between the sthdnin and the sthdna is reversed, and the genitive as defined by A 1.1.49 may be said to identify the sthdnin as that linguistic element which would possibly appear if a certain rule did not introduce the ddesa. This presupposes that the sthdnin and the ddesa are in complementary distribution. In the case of A 2.4.52 aster bhuh, then, the genitive serves to identify as which is the element that would possibly appear if the rule did not specify the element bhu as its replacement in the domain of drdhadhdtuka suffixes. With the above background it is possible to investigate more closely the relation marked by the genitive as defined by A 1.1.49 in the light of how

Substitution

239

Paninlyas themselves have considered the issue. Kaiyata, in his commentary on rule 1.1.49, remarks on the relation in question as follows (MbhP P.408b): sthane yogo 'sya iti I sthdnammittasambandhety arthahll trtlyaya vefi/ 'nipdtandf iti yojyaml yathd devadattasya yajnadatta iti putratvddidvdrakah sambandho 'vagamyate, evam aster bhur ity atrdpy asydm paribhdsdydm satydm sthdnanimittah sambandho 'vasiyate II 'That of which there is a relation in the sthdna': that is to say, [the genitive identifies] a relation which is caused by the sthdna. [On] 'trtiydyd va? [one should add that this statement too] is to be constructed with nipdtandt 'because it is a legitimate form'. When you say 'Devadatta's Yajnadatta' a relation that becomes apparent through such properties as 'being a son' etc. is understood. Likewise with regard to the statement 'aster bhuh\ given this principle of interpretation,153 a relation which is caused by the sthdna is ascertained too. Kaiyata here identifies yoga as sambandha 'relation', and the way in which Patanjali analysed the compound sthdneyogd makes him say that the relation in question is one which has the sthdna as its cause (nimitta). To tell us how relations are perceived, he offers the by now familiar example of devadattasya yajnadattah, where the property of putratva 'being a son' or some similar abstract relational property has to be supplied if the relation between Yajnadatta and Devadatta is to be understood. If the paribhdsd, the principle of interpretation expressed by rule 1.1.49, is recognised, the relation expressed in rule A 2.4.52 aster bhuh is understood through sthdnatva in the same way that the relation marked by the genitive in devadattasya yajnadattah is established through putratva, the property of being a son, or any similar abstract relational property. Given the paribhdsd rule A 1.1.49, rule 2.4.52 aster bhuh becomes understood as well. On the available evidence so much seems to be Kaiyata's view. Now, it is well known that Kaiyata built heavily on Bhartrhari. Consider then the following passage from Bhartrhari's DTpikd on rule 1.1.49 (MbhD 288,17-289,2): darsanadvayam ca sasthinirdesesu I 'aster bhuh' iti I eke manyante154 aster iti yd sasthi sd naiva bhuvah sambandhinil kim tarhi sthdnaprayuktd I ddheyatvena tu bhavatir upddiyate iti I ke cid dhi sabdd na svarupdd vyatirekam upajanayitum saknuvantil tad yathd I devadattasya yajnadatta iti na yajhadatto yajhadattarUpdd devadatte155 vyatirekam janayati I evam hi kalpyamdne samuccayapratipattih sydt - devadatto yajhadattas ceti1561 tasmdd ye ynye yrthd devadattasya yajnadattah pita putrah svdmT bhrtyah samipa ity etebhyo vyatirekah1511 ke cit tu svarupdd eva
153 154 155 156 157

That is, rule 1.1.49 itself. So phrase with Palsule and Bhagavat (1991:19,7-8) for darsanadvayam sasthinirdesesu 'aster bhuh? ity eke manyante I. So MbhDms fol. 201a, a reading retained by Palsule and Bhagavat (1991:19,11), for MbhD devadatteti. I am not fully confident of the text or my translation here. So also Palsule and Bhagavat (1991:19,12). The MbhDms fol. 201a reads yajnadattasyeti. So also Palsule and Bhagavat (1991:19,13). The MbhDms fol. 201a reads vyatirekecittu0.

240

Indian semantic analysis

vyatirekam dhur itil yathd ddsahputrah karmakara itil anye manyante Ibhusabda evaparigrhya svasamavdyino dharmdn sdmipyddin vyatirekahetur15* itil Now there exist two views concerning statements where genitives are given, such as aster bhuh. Some hold that the genitive in asteh is not a relatum of bhuh at all, but rather it is linked directly to the sthdna. [The verb] bhavatih (^bhu), however, is employed as something to be placed [in the sthdna]. For some linguistic units are not able to produce additional relational difference on the basis of their own form, just as in the statement 'Devadatta's Yajnadatta' where [the term] Yajnadatta does not produce any relational difference on the basis of the form 'Yajnadatta' with regard to Devadatta. For when it is considered like this, we get a mere understanding of aggregation, namely, 'Devadatta and Yajnadatta'. Therefore it is from those other senses, such as Yajnadatta [who is] father, son, master, servant, [or] near to Devadatta, that a relational difference [is produced]. Some [linguistic units] do, however, express a relational difference on the basis of their own form alone, such as 'slave', 'son', 'employee'. Others hold that the linguistic element bhu itself, once it has taken on properties such as nearness which are inherent in it, is the^use_with-i:egard to [producing] relational difference. On the authority of Bhartrhari, then, there exist two positions concerning statements which contain genitives to be interpreted according to A 1.1.49. Such a statement is A 2.4.52 aster bhuh, which, taken by itself, clearly contains a relational genitive and involves a relation between two relata, the elements as and bhu, however it is impossible to know from this statement alone just what exactly the relation in question is. In this respect the statement aster bhuh is similar to devadattasya yajnadattah, where the term yajnadatta, just like the term bhu, is incapable of producing any relational difference on the basis of its own form since it does not belong to the class of inherently relational terms such as pitr 'father', puira 'son', svdmin 'master', etc. Thus, all that is achieved by a statement like devadattasya yajnadattah is an understanding of mere aggregation, that is, one entity and another entity. If, on the other hand, one of the two terms is a relational term like pitr and such, the relation in question is immediately understood; for example, devadattasya pita 'Devadatta's father'. Now, if one adds the locative grhe 'in a house' to form a sentence like devadattasya grhe yajnadattah, one immediately understands that Yajnadatta is in a house that belongs to Devadatta. The genitive devadattasya is directly bound to grhe. Similarly, if one supplies sthdne to the statement aster bhuh to give asteh sthdne bhuh, one immediately understands as to be related directly to what sthdne signifies. The term bhu, on the other hand, since it is not a relational term such as pitr 'father' or anything similar, simply refers to the verb. In context, this is referred to as something which is to be put in a place. This indicates that Bhartrhari considers sthdna as a term referring to a locus.
158

So read with MbhDms fol. 201a, a reading retained by Palsule and Bhagavat (1991:19,16), for MbhD vyatireke hetur in.

Substitution

241

The other position alluded to in the DTpikd passage under discussion is that even such terms as bhu or yajnadatta, although they are not inherently relational terms such as pitr 'father', nevertheless can serve to show additional properties. In this case bhu is directly related to as and able to produce relational difference by itself once it is qualified by a relational property. Now, given the paribhdsd rule sasthi sthdneyogd which states that a genitive which is not already contextually bound is to be understood as involving a relation with the sthdna, once the locative word form sthdne is supplied in accordance with this provision, a genitive such as asteh is indeed linked directly to sthdne: asteh sthdne bhur bhavati, 'bhu applies in the sthdna of as\ Let me now turn to Nagesa's commentary on A 1.1.49 and the way in which he remarks on the relation denoted by the genitive as defined by this rule (MbhU I:408b): atra sthdnasdmipyddayah sasthyarthasambandhanirupakdh, na tu sasthyarthd iti na matupsutrasthabhdsyavirodhah I ata evdtra bhagavatd saptamyantena trtiydntena vd samdso darsitah, na tu samdnddhikaranabahuvnhih I tad dhvanayatoktam - sthdnanimittasambandhety artha itiII nanv ddesavidhau sthdnasabddbhdvena katham sthdnanimittah sambandho 'ta dha-yatheti I tathd ca 'devadattasya yajhadattaK ity ukte kim ity dkdnksdvasdt putra ity adhydhdro yathd, tathd pradesesu sthdnapadddhydhdrena vdkyaikadesanydyena vd bodha iti bhdvahll There is no disagreement here with what is stated by the Bhdsya in the context of the matUP rule since [terms] such as sthdna and sdmipya condition the relation which is the meaning of the genitive case, but they are not the meanings of the genitive case. For this very reason Patanjali presents this as a compound with a word ending in the locative or the instrumental case, not as a bahuvnhi compound whose members are in syntactic agreement. So to bring this out [Kaiyata] has stated: 'that is to say, it is a relation which is caused by the sthdna\ But since the word sthdna is absent in rules which teach substitutes, how [do we know] that the relation is one which is caused by the sthdna? For this reason he [= Kaiyata] makes the statement 'just like'. So in this way, when it is stated 'Devadatta's Yajnadatta', by force of the expectancy [which is reasonably expressed by the question] 'what?', just like you supply putra 'son', in the same way you get an understanding by supplying the word sthdna in the rules [which constitute the range of its application] (pradesesu), or by the principle whereby part of a sentence [brings out the meaning of the whole sentence], that is the idea. Nagesa here agrees to the interpretation of the compound sthdneyogd as a vyadhikaranabahuvnhi, a bahuvnhi compound whose members are not in syntactic agreement. The reason for this interpretation is, according to Nagesa, that sthdna, sdmipya 'nearness', and dnantarya 'adjacency' are not meanings of the genitive case. Rather they exhibit various relational properties which are necessary conditions for the relation marked by the genitive case, like Kaiyata also said. As mentioned already,159 Vasudeva Dlksita fixes the number of such terms
159

See p. 214 above.

242

Indian semantic analysis

at three in his Bdlamanoramd (SK 1:44,4-5): sabdasya sabdena traya eva sambandhdh - dnantaryam sdmipyam prasangas ca, 'only three relations [may obtain] between a linguistic unit and [another] linguistic unit: adjacency, nearness and possible appearance'. He, however, refers to these three terms simply as relations and takes sthdna in the sense of prasanga 'possible appearance'. Nagesa spots a difficulty in his own interpretation and is careful to note that it is not in conflict with views stated by Patanjali in the context of the matUP rule. It may be clarifying to examine this briefly. The rule referred to is A 5.2.94 tad asydsty asminn iti matup, which introduces the possessive suffix matUP (unaccented -mat) to form derivatives in the senses of 'that is pertaining to it' or 'that is located in it'. This rule comes under the domain of the heading rule A 4.1.82 samarthdndm prathamdd vd which states that rules introducing derivative suffixes called taddhita apply optionally and that the suffix introduced by each rule is added after the first of the semantically and syntactically related (samartha) elements referred to in each rule. In a rule such as A 5.2.94, then, the suffix matUP is introduced after the variable tat 'that', which refers to a nominative form denoting aijjobject, in the senses of 'asydstV or 'asminn asti\ Or, speaking with the Kdsikd,160 asmin 'in it' and asya 'pertaining to it' are the senses of the suffix, while asti 'is' is a qualifier of the stem. An example of such a derivative is gomat 'possessing cows; rich in cows'. According to rule 5.2.94 this is semantically equivalent to gdvo 'sya santi, 'cows are his'. In his discussion of A 5.2.94 Patanjali brings up the following interpretation of the relation denoted by a genitive case suffix (Mbh 11:391,17-23): idam tarhi prayojanam astiyuktdd yathd sydd anantarddiyuktdn md bhud iti I gdvo 'sydnantardh gdvo 'sya samipa iti I atha kriyamdne }py astigrahana iha kasmdn na bhavati gdvo 'sya santy anantardh gdvo 'sya santi samipa iti I asdmarihydt I katham asdmarthyam I sdpeksam asamartham bhavatitil yathaiva tarhi kriyamdne 'stigrahane 'sdmarthydd anantarddisu na bhavaty evam akriyamdne 'pi tia bhavisyatil asty atra visesahl kriyamdne 'stigrahangjtdntarena trtlyasya padasya prayogam anantarddayo 'rthd gamyante I akriyamdne punar astigrahane 'ntarendpi trtiyasya padasya prayogam anantarddayo 'rthd gamyante II This, then, is the purpose [of rule 5.2.94], namely that it (= the suffix matUP) should be added after that which is connected with asti, and that it should not be added after that which is connected with anantara etc. such as gdvo 'sydnantardh, 'cows adjacent to him', or gdvo 'sya samipe, 'cows in the nearness of him'. Now, even when asti is mentioned, why does it not apply here: gdvo 'sya santi samipe, 'there exist cows adjacent to him', gdvo 'sya santi samipe, 'there exist cows in tlie nearness of him' ? The reason is that there is absence of semantic connection. How [can you say that there is] absence of semantic connection? [Because of the principle saying:] 'that which requires [qualification by a co-occurrent word] is [regarded as] semantically unconnected.' But this amounts to saying that when mention of asti is made, it does not apply with regard to anantara etc., and that likewise it will not apply
160

asyasminn iti pratyayarthau I astiti prakrtivisesanam I

Substitution

243

when [mention of asti] is not made? There is a difference here. When mention of asti is made, the senses of anantara etc. are not understood unless some third word is employed [in which case there would be lack of semantic connection]. But when mention of asti is not made, the senses of anantara etc. are understood even without employment of some third word. The statement sdpeksam asamartham bhavati, referring to the principle that an element of a formation which requires some outside element as its own qualifier is regarded as semantically unconnected and therefore unsuited for the formation in question, is commented upon by Kaiyata in the context of A 2.1.1 samarthah padavidhih. This rule stipulates that any operation concerning finished words (pada) applies only to linguistic elements which are semantically connected. Although the context here is the formation of words into compounds, the basic principle is the same in the formation of words by taddhita suffixes. In either case semantic connection has to hold. Kaiyata explains the principle in question as follows (MbhP II:318b): nydyo 'yam I vrttau hy upasarjanapadena pradhdndrthdbhidhdyind bhdvyaml svavisesandpeksdydm ca tasya prddhdnyam iti katham ekasyaikadd prddhdnyam ekdrthibhdvas ca sydtll This is a maxim. For, in formation [of compound words], the subordinate word should denote the meaning of the principal word [of the formation]. But when [the subordinate word] requires [some third co-occurring word as] its own qualifier, then principality pertains to it.161 How, then, can one [word] at any one time be principal and how can there be a meaning-unification?162 Returning to rule 5.2.94, we meet with the following statement in Nagesa's commentary on the Bhdsya discussion quoted above (MbhU IV:343b): 'astind yuktdd evcC iti vaktum ucitam, 'atha kriyamdne 'pi' ity uttarabhdsyasvarasdt, 'it is proper to state "after that which is connected with asti alone" because of the inclination later in the Bhdsya [to state]: "now even when [mention of asti] is made'". Consider now the following passage from Kaiyata's remarks on the Bhdsya discussion of A 5.2.94, keeping in mind the view presented in the Kdsikd that the word asti in rule 5.2.94 is a qualifier of the base (MbhP IV:343b-344a): kriyamana iti I prakrtyarthavisayapaddpeksam trtiyatvam - gdvah santy anantara asyetil tatrdstigrahane kriyamdne kevala evdstyartha upddhibhdvena pratyaydnte 'ntarbhavati/ asati tu astigrahane dharmdntardvacchinnasydpy astyarthasydntarbhdva dsankyetety arthahl tatra kriyamdne 'stigrahane 'styarthasyaivdntarbhdvdvagamdd vdkye yat trtiyam padam - 'anantardK iti tad vrttau svdrthapratipddandydvasyam prayoktavyam I tatprayoge cdsdmarthyaml akriyamdne tv astigrahane 'nekasydpy upddher antarbhdvasambhdvanayd vrttdv anantarapadaprayogam antarendpy dnantarydvagamdd asdmarthyam na sydd ity arthahlI
161 162

That is, to the subordinate word with regard to its own qualifier. That is to say, the original pradhdna and the original upasarjana cannot both be the pradhdna at one and the same time.

244

Indian semantic analysis

'When [mention of asti] is made': being a third [word] involves a relation of dependence with regard to a word that is within the range of the stem's meaning, such as in gavah santy anantara asya, 'there exist cows adjacent to him'. As lar as this is concerned, when asti is mentioned, the meaning of asti by itself is included in that which ends with the suffix as an extraneous [meaning-]condition [of the stem]. But when there is no mention of asti, one may suspect the inclusion even of the sense of asti specified by these other features; that is the sense. In this respect, when mention of asti is made, that which is the third word, namely anantardh, must necessarily be used in order to convey its own meaning in the formation, since in the sentence we understand inclusion of the sense of asti only. And when it is used, there is absence of semantic connection. But when mention of asti is not made, through the possibility of including even diverse extraneous [meaning-]conditions in the formation, no lack of semantic connection would obtain, because there is understanding of dnantarya 'adjacency', even without the employment of the word anantara 'adjacent'; that is the sense. Without entering into a detailed discussion of theintricacies involved in the interpretation of A 5.2.94, it is clear from tlje-abbve remarks that if a suffix A is applied to a stem B, semantic connection has to hold between A and B. That B is qualified by some third word C, such as anantara, means that B is disconnected from A, so that meaning-unification, ekdrthibhdva, dobs not obtain. In the example here, the word gavah 'cows' is qualified by the word anantardh 'adjacent' which is related to gavah alone. Such a dependence on an additional word precludes the formation of gomat-, since anantardh is a qualifier of the meaning of gavah, that is, of the stem's meaning, which does not extend to the meaning of the suffix conveying the sense of the genitive case suffix. Thus, when Nagesa in his remarks on A 1.1.49 claims that there is no disagreement between what Patanjali says in the context of A 1.1.49 and what he says in the context of A 5.2.94, this provides us with further evidence that sthdna, sdmipya or dnantarya are not meanings of the relation marked by the genitive case suffix. When Patanjali at A 5.2.94 talks about lack of semantic connection, asdmarthya, this is because he regards anantara 'adjacent' as a qualifier of the stem's meaning, while the genitive variable asya is the sense of the suffix that is introduced. The term sthdna, then, is not the meaning of the genitive case, and it is accordingly not included among relational terms, that is, it is not a member of the class of all possible relations. On the contrary, the term*sthdna itself determines the relation marked by the genitive case suffix so that the particulars of the relation are understood. This is the view presented by Nagesa also in his Laghusabdendusekhara (LSI:58,l-2): sthdnapaddrthanirupitasambandhdrthikety arthah, 'that is to say, [sthdneyogd signifies] something which has as its meaning a relation determined by the sense of the word sthdna\ Rule 1.1.49, then, basically teaches that a genitive is used when the relation marked by it is determined by the locative word form sthdne. There remains the problem that the word sthdne is absent in rules which teach substitution operations. Kaiyata, in the passage quoted above, states that

Substitution

245

there is the understanding of dnantarya 'adjacency' even when the word anantara 'adjacent' is not used. Similarly, Nagesa, in the passage quoted above from his commentary on A 1.1.49, states as a preliminary solution that just like one applies, by expectancy, some further qualification like putra 'son' on hearing a statement such as 'Devadatta's Yajnadatta', in the same way one gets meaning by supplying the necessary element, that is, by supplying the word form sthdne in rules which are within the range of application of A 1.1.49. Nagesa also presents an alternative solution. Since the definition in the rule is sasthl sthdneyogd, then, whenever you hear a sasthl, that is, whenever you hear a genitive case suffix in rules for the interpretation of which A 1.1.49 must come into play, you supply the rest of the rule, namely sthdneyogd, on the principle of vdkyaikadesa,163 the principle that part of a sentence brings out the meaning of the whole sentence. The problem then is to decide the domain of A 1.1.49, that is to say, to decide when a genitive case suffix in a rule is to be interpreted according to rule 1.1.49 and when it is not. Concerning the interpretation of the relation marked by a sthdnasasthT, some important remarks are also given by the commentators in the context of A 6.4.1 angasya. This is a heading rule, adhikdra, recognised by the tradition as valid throughout the seventh adhydya. Basically, rule 6.4.1 states the condition that the name anga 'presuffixal base' must apply if the operations signalled by the rules within this section are to take place. The interpretation of the genitive angasya thus affects several hundred rules. Now the interpretation of this genitive raises certain difficulties, details of which need not concern us here.164 But a brief summary of what Katyayana and Patanjali say seems necessary. The discussion of the genitive at A 6.4.1 begins by assuming that angasya is a sthdnasasthT, that is, a genitive as defined by A 1.1.49. To this the objections are raised that one must also have a heading with an ablative form, arigdt, and that such forms as partitive genitives in the angddhikdra are not properly accounted for.165 These objections are answered by saying that the class names pratyaya 'suffix' and anga determine each other - that is, that an anga requires a suffix and a suffix an anga - and that angasya contains a genitive signifying a relation.166 Thus, for example, the relational genitive angasya is understood in a rule such as A 7.1.9 ato bhisa ais, which teaches that ais is
163 164 165

On this nydya see Mbh 1:111,22 f. For a full discussion of this rule, see Benson 1990:72-123. A 6.4.1, vt. 1: angasyeti sthdnasasthT cet paficamyantasya cddhikdrah, 'if the genitive angasya is a sthdnasasthT, then another heading of the word ending m the ablative [has to be stated]', and vt. 2: avayavasasthyddindm cdprasiddhih, 'moreover, the genitives which mark a part etc. are not properly accounted for'. See Benson 1990:87 ff. 166 A 6.4.1, vt. 3: siddham tu parasparam pratyangapratyayasamjndbhdvdt, 'but it is prope accounted for because the technical names anga and pratyaya are mutually dependent', and vt. 4: sambandhasasthTnirdesas ca, 'and [this is] a statement of a genitive which denotes a relation [in general]'. See Benson ibid. The term anga itself is defined by rule 1.4.13 yasmdt pratyayavidhis tadddi pratyaye 'ngam as 'that which begins with that [linguistic item] to which a suffix is added and is followed by the suffix'. See Benson 1990:13-71 and p. 201, note 60 above.

246

Indian semantic analysis

substituted in the place of the instrumental plural suffix bhis when it follows a short a, so that this rule does not wrongly allow ais to replace bhis of an item like bhissd 'cooked rice' in brdhmanabhissd 'Brahman's rice'. The substitution is stated for bhis related to an anga, that is, which follows a short a of an element to which the name anga applies, so that the bhis in question must be a suffix (pratyaya). As Kaiyata remarks in this context, an interlocutor who does not understand the full impact of this assumes that only one relation is intended, the one involving a determining condition and a conditioned element (nimittanimittibhdvasambandhah) instead of any possible relation, depending on the context in which the heading angasya is allowed to recur. At this point, then, Patanjali brings up the objection that genitives such as partitive genitives are not included, and that the sthdna too is not included when the heading s angasya recurs in subsequent rules (Mbh III: 179,7-8): evam apy avayavasasthyddayo 'visesita bhavantil avayavasasthyddayo 'pi sambandha evall evam api sthdnam 'visesitam bhavati Ijthdnam api sambandha evall Even so the genitives which mark a part-^fcTremain unqualified [by the word angasya]. Also the genitives which mark a part etc. exist only when there is a relation. Even so the sthdna remains unqualified. The sthdna too is precisely a relation. This passage raises a problem created by the ambiguous sandhi of sambandha eva, that is, whether the case form intended is the locative, sambandhe, or the nominative, sambandhah. Taking into consideration the parallelism of the two statements, avayavasasthyddayo }pi sambandha eva and sthdnam api sambandha eva, one would immediately feel inclined to interpret them in the same way. Were this justified the latter statement would translate: /the sthdna too exists only when there is a relation'. However, I do not think this is the case. When Patanjali considers the avayavasasthT and such, saying of these sambandha eva, commentators do indeed accept the form to be sambandhe, locative. If we take avayava quite literally as 'part', then we have to accept one interpretation, suggested by Nagesa {MbhU V:270b): avayavasasthyddayo 'pi sambandhe eva -jdyanta ity arthah, 'that is to say, even the genitives which mark a part etc. arise when a relation is to be signified'. If, on the other hand, one considers avayava to be a metaphorical usage of a relational property, that is to say avayavatva 'part-ness', then one can say, with Nagesa, that in the end these abstract relational properties are included among relations in general (MbhU V:270b): yad vd tadvdcyd avayavatvddayo167 'pi sambandhe 'ntarbhutd ity arthah, 'or else, even part-ness {avayavatva) and such are to be expressed by it, that is to say, these are included in [the notion of] relation'. Thus, according to Nagesa, such genitives as the avayavasasthT can either be conditioned by a relation or be included among abstract relational properties, that is, among relations in general. ' Now, given that what is considered to recur in rules headed by A 6.4.1
167

Vedavrata (1962-3.IV 665) reads tadvdcydvayavatvddayo0'.

Substitution

247

angasya is merely this term which is to be interpreted contextually, one has only a relational genitive. The particular relation in question is either given contextually or, if this is not the case, provided for by A 1.1.49 sasthT sthdneyogd. Thus, for example, sdsah of A 6.4.34 sdsa id anhaloh is understood to contain a partitive genitive, since the word is contextually bound to upadhdydh which recurs from A 6.4.24 aniditdm hala upadhdydh kniti. And the genitive angasya, which recurs from A 6.4.1 and is co-referential with sdsah, is also thus bound and accordingly to be interpreted as a partitive genitive. Now sdsah is understood to recur in A 6.4.35 sd hau where it is not a bound genitive, and so the co-referential genitive angasya recurring from rule 6.4.1 is likewise not bound, so that in this case one understands sdsah angasya sthdne . . . 'in the sthdna of sds that is an aftga . . .' by rule 1.1.49 sasthT sthaneyogd}^ In the light of the above, consider again Patanjali's statement sthdnam api sambandha eva, the answer to the objection brought up in the Bhdsya that the sthdna would remain unqualified by the term angasya in rules where the latter recurs if angasya merely contains a relational genitive. Now, the partitive genitives and such could be conditioned by or included among relations. But the term sthdna, if it means prasanga 'possible appearance; chance' or artha 'meaning', cannot be included among relations because it is not itself a relation, nor can it be conditioned by a relation since it itself determines a relation. Still I think it pretty certain that sthdnam api sambandha eva is indeed to be understood, with commentators, as sambandhah eva, not as sambandhe eva. Kaiyata comments on the statement in question as follows (MbhP V:270a): sambandhanimittatvdt sambandha ity arthah, 'the meaning is that by fact of its being a necessary condition for a relation, it is a relation'. Referring to the two interpretations of the term sthdna just mentioned, prasanga and artha, Nagesa adds the following explanation to Kaiyata's remarks (MbhU V:270b): nanu prasango 'rtho vd sthdnam tanmrupito hi sambandho na tu tad eva sambandho 'ta aha - sambandhanimittatvad in / sambandhinau hi sambandhasya nimittaml nimittanimittinor abhedopacardd evam uktir ity arthah// Now sthdna signifies either possible appearance or meaning. However the relation is determined by it, the relation is not it alone, and therefore he says: 'because it is a necessary condition for a relation'. For its two relata are the necessary condition for a relation. The sense is that since there is metaphorically no difference between a condition and that which is conditioned [by it], such a statement [can be made].
168

For the example, cf Cardona 1974a 310 Rule 6.4.34 sdsa id anhaloh provides for substitution of the vowel i for the penultimate sound (upadhdydh, recurring from A 6 4.24) of the verbal root Vms1 'order; instruct' when the latter is followed by aN or by any suffix beginning with a consonant marked by a diacritic KoiN {kniti, recurring from A 6 4.24) Rule 6 4 35 sd hau signals that the presuffixal base sds is replaced by sd before the second singular imperative ending hi to give sddhi (cf Cardona 1970a.49, with note 36)

248

Indian semantic analysis

Similarly, the seventeenth-century commentator Annambhattajstates in his Uddyotana on Kaiyata's Pradipa (ed. Narasimhacharya 1982:269): sthdnyddesabhdve sthdnam nimittam ity upacdrdt 'sthdnam sarribandhaK ity uktam, na tu sthdnam eva sambandha ity aha - sambandhanimittatvad itil When there is a substituend-substitute relation, the sthdna is a necessary condition, and so, by transfer of denotation,169 it has been stated that the sthdna is a relation. But the relation is not the sthdna alone, and so he says: 'by fact of its being a necessary condition for a relation'. So, just as the relation expressed by the statement devadattasya yajnadattah, 'Devadatta's Yajfiadatta', is conditioned by some other term such as putra 'son', in the same way a relation expressed by the words aster bhuh is conditioned by the locative word form sthdne, although the situation of the sthdnasasthT is slightly different from that of the avayavasasthyddayah, the genitives which mark a part, etc. The latter could, as Nagesa points out, be conditioned by a relation or be included among relations, that is, be members of the class of all possible relations. On the other hand, the sthdna, when the term signifies prasanga 'possible appearance; chance' or artha 'meaning', is not included in the class of all possible relations, because it is not itself a relation, nor is it conditioned by a relation. On the contrary, the sthdna itself determines a relation. Hence I think it proper to accept what Kaiyata and Nagesa say, which also Annambhatta says: the sthdna is called a sambandha 'relation' metaphorically, because it is a necessary condition for the relation in question. Note that I have accepted what most Paninlyas say, namely that the sthdna is not itself a sambandha and thus not something directly conveyed by a genitive ending, but it is a necessary condition for the relation between the sthdnin and the ddesa and hence metaphorically considered a relation. The meaning of the term sthdna It is now possible to investigate more closely the meaning of the term sthdna. As mentioned already, this is basically analysed in two ways by the commentators. Firstly, it may be interpreted as bhdvasddhana, that is to say* as an action noun formed with the suffix LyuT (-ana with presuffixal accent) by A 3.3.115 lyut ca. Under this alternative it denotes mere activity, state of being, equivalent to sthiti 'a standing'. So, for example, Jinendrabuddhi's Nydsa on the Kdsikd at A 1.1.49 (KasN 1:168): sthdnasabdo bhdvasddhanah sthitih sthdnam, 'the word sthdna [is a noun] realising mere activity, sthdna [thus] being equivalent to sthiti "a standing'", or Nagesa's Laghusabdendusekhara (LSI 58,6):
169

Strictly speaking, upacdra does not imply exactly the same as the Western*notion of metaphor. Rather, upacdra means that a primary concept A is denoted not by the usual a but by an accidental or secondary b, that is to say, it involves a change of denotation. See in this respect G. Gren-Eklund 1986. Thus, in the present case, the word sambandha is substituted for sambandhanimitta.

Substitution

249

bhdve lyut sthdnasabde, 'LyuT in the sense of mere being in the case of the word sthancC. Alternatively, it is interpreted as adhikaranasddhana, that is to say, as a noun denoting a locus, in this case formed with LyuT by A 3.3.117 karanddhikaranayos ca. So, for example, Kaiyata on the Mahdbhdsya at A 1.1.1 and A 1.1.49 (MbhP 1:162a, 408b): tisthanty asmin sabddh, 'speech elements stand in it'.170 Under the first alternative commentators remark that three particular contextual meanings can be considered for the word sthdna: apakarsa 'drawing away', nivrtti 'ceasing; cancellation', and prasanga 'possible appearance; chance'. Thus, for example, the Nydsa on the Kdsikd at A 1.1.49 (KasN 1:168; similarly PadamahjarT, ibid.): gosthdne 'svo badhyatdm ity apakarsah sthdnasabdasydstheyah I slesmanah sthdne kadukam ausadham ity atra nivrttihl darbhdndm sthdne sarair dstaritavyam iti prasangahl When one says 'Let a horse be tied up in the sthdna of the cow', apakarsa 'drawing away' is to be recognised for the word sthdna. When one says 'Bitter herb medicine in the sthdna of phlegm', here nivrtti 'ceasing; cancellation' [is to be recognised]. When one says 'One should strew with sara in the sthdna of darbha grass', prasanga 'possible appearance; chance' [is to be recognised]. This amounts to the following. When one says 'A horse should be tied up in the place of the cow', one means that the cow in question is to be taken away and a horse tied up in its place; when one says 'Bitter herb medicine in the case of phlegm', one means that the phlegm is removed when the medicine is taken; and when one says 'One should strew with sara in the place of darbha grass', one means that if the occasion arises that darbha grass is not available, then one uses sara instead. That is to say, the particular senses in question are contextual varieties of a single general meaning of the word sthdna. In the context of A 1.1.49 sasthi sthdneyogd Paninlyas generally take sthdna in sthdneyogd to be an action noun in the sense of prasanga 'possible appearance; chance'. For instance, A 1.1.51, vt. 3: siddham tuprasange raparatvdt, 'but it is properly accomplished because a following r is taught when there is the possible appearance [of aN in the sthdna of r]', 171 with Bhdsya (Mbh 1:126,15-17):m uh sthdne 'nprasajyamdna eva raparo bhavatitiIkim vaktavyam etatlna hi!katham anucyamdnam gamsyatel sthdna iti vartate sthdnasabdas ca prasangavdcTI
170

171 172

So also Bhartrhari in his DTpikd (MbhD 289,2-9); a full discussion of the relevant passage will be given below. Note that this analysis is a nirvacana analysis. The question arises as to whether the commentators apply A 3.3.115 and 117 on purpose or mechanically. In other words, do they apply these rules because they consider them to involve an important distinction or because there exist two rules by which the term may be analysed? According to A 3.3.117 they could also have applied an analysis as karana, an option they do not avail themselves of. Rule 1.1.51 ur an raparah teaches that a sound denoted by aN replacing r is followed by r. Cf. also A 1.1.52 with Mbh 1:130,14.

250

Indian semantic analysis

r applies only when aN is possibly appearing in the sthdna of r. Is this to be stated? Indeed not. [But if] it is not to be stated, how is it to be understood? [The word form] sthane is still current, and the word sthdna signifies prasanga 'possible appearance'.173 Under the second alternative, the interpretation as adhikaranasddhana, one particular locus that is considered is the artha 'meaning' of a linguistic element or speech unit, sabda. Thus, for example, Kaiyata on the Mahdbhdsya at A 1.1.1 (MbhP 1:162a): sthdnasabdo 'rthavdcT tisthanty asmih sabdd iti sthdnam, 'the word sthdna signifies artha "meaning" [according to the analysis]: "something occurs [tisthanti, lit. 'stands'] in it, namely linguistic elements", thus [it is called] sthdna\ These two interpretations of the term sthdna will now be investigated in the light of what the commentators remark in the cornet of A 1.1.49. Bhartrhari on sthana ,"

There is an important passage in the DTpikd which deals with the interpretation of sthdna as artha where Bhartrhari brings forward the argument that substitutions are of whole units. Moreover, Bhartrhari mentions here still another possibility, based on the position that all linguistic elements are first selfreferring (svarupapaddrthaka), a position alluded to already by K^tyayana.174 Thus, let us assume that the sthdnins referred to by as in A 2.4.52 aster bhuh are instances of as actually used to signify meaning. The meaning of these, then, is signified by bhu. Now, the pratydhdra115 terms iK, yaN, and so on are technical terms (sabdasamjhd) which signify i and so on, and the vowels ultimately signified also have meanings, namely their own forms (svarupa). Even these sounds, then, can be considered to have meanings, the sounds that occur in actual terms such as dadhi, dadhy. '*. The DTpikd passage in question is indeed not simple, but then so much of this text is badly transmitted and difficult. Since this is a passage of considerable importance, revealing as it does the views of Bhartrhari, one of the most
173

174

175

Further textual evidence is easily adduced. In addition to the Kdsikd, Nydsa, and Padamanjan on A 1.1.49, see for example the Sabdakaustubha (SK 191) and Siddhdntakaumudl (SK 1:44) of Bhattoji Diksita {sthdnam ca prasangah), and also two commentaries on tfre latter work, the Bdiamanoramd of Vasudeva Diksita and the Tattvabodhim of Jnanendra SarasvatI (SK 1:44-5). See also the Jainendramahdvrtti (JV) of the Jain author Abhayanandin on the corresponding rule 1.1.46 td sthane of Devanandin's Jainendravydkarana. (Instead of referring to a case ending, vibhakti, in the customary way of numbering them with the ordinals first through seventh, Devanandin uses the device of adding a to each letter of the Word v-i-bh-ak-t-i. A sixth triplet ending, sasthi, is thus referred to as td.) A 1.1.68 vt. 2. Rule 1.1.68 svam rupam sabdasydsabdasamjhd states that a linguistic element denotes its own form except in the case of a technical term, sabdasamjfid, given in the grammar. Vt. 2 states: na vd sabdapurvako hy arthe sampratyayas tasmdd arthanivrttih, not, for the conceptualisation of meaning is preceded by the knowledge of linguistic elements, and therefore there is cancellation of meaning'. A pratydhdra term is a technical abbreviation denoting a set of sounds, iK thus denoting long or short iurl.

Substitution

251

prominent among all the Indian grammarians and philosophers of language, I shall deal with it in considerable detail. As is well known, Bhartrhari's Dipika is extant in a single fragmentary manuscript which, moreover, is written by various hands and is full of corruptions. To begin with I quote the text as it was adopted by the editors of the first printed edition of the manuscript, K.V. Abhyankar and V.P. Limaye (MbhD 289,2-9): sthdnam cdrthahl tisthaty asmin sabda iti asmin aster arthe bhavati I bhdvini vrttis ca itarasya yasmdd bhavati tadartham dhetil atha varnddesesu kah sthdndrtho ydvatd 'rtha eva varndndm ndstil tatrdpi 'sarve sarvapadddesdK ity asmin pakse dadhisabdasya dadhy ddeso ruher api rupih iti yuktaml atha vd kdlah sthdnasabdenocyate I ikah kale yano 'dhitisthanti iti/ yathd vd 'asteK ity etad anukaranam arthaprayuktdndm asirupdndml te ca sthdninahl tesdm artho vidhvyate bhavatindl evam iko yan acity atrdpi iko yani krte dadhy ddih sydt, ikah tesdm yat svarupam so 'rthah sthdnam iti II G.B. Palsule and V.B. Bhagavat (1991:19,17-26) suggest the passage be read in the following way: sthdnam vdrthah, tisthaty asmin sabda iti I asmin \pakse] aster arthe bhavatir bhavati I nivrttis cetarasya yasmdt bhavatis tadartham dheti I atha varnddesesu kah sthdndrtho, ydvatd artha eva varndndm ndstil tatrdpi 'sarve sarvapadddesdK ity asmin pakse dadhisabdasya dadhy ddeso, ruher api rupir iti yuktaml atha vd, kdlah sthdnasabdenocyate - ikah kdlam yano 'dhitisthantiti I yathd vd 'aster' ity etad anukaranam arthaprayuktdndm asirupdndm, te ca sthdninah tesdm artho 'bhidhiyate bhavatind, evam Hko yan acV ity atrdpi iko ye 'nukrtd dadhyddisthd ikas tesdm yat svarupam so 'rthah sthdnam iti II At several points here I differ from all the editors in the text adopted. On the basis of the actual readings found in the DTpika manuscript (MS = MbhDms),1761 propose the passage be read as follows: sthdnam cdrthahl tisthaty asmin111 drsta iti I asminn aster arthe bhavatihl bhdvinivrttis cetarasya yasmdd bhavatism tadartham dhetil atha varnddesesu kah sthdndrtho ydvatdrtha eva varndndm ndstil tatrdpi sarve sarvapadddesd119 ity asmin pakse dadhisabdasya dadhyddeso ruher api rupirm iti yuktaml atha vd kdlah sthdnasabdenocyate I ikahm kdlam yano 'dhitisthantiti I yathd vaster ity etad anukaranam arthaprayuktdndm1*2 asirupdndmml te ca sthdninahl tesdm artho vidhiyate bhavatindl evam iko yan acity atrdpy iko ye na krtd dadhyddisthdm ikas tesdm yat svarupam so 'rthah1*5 sthdnam iti I On the basis of this reading I offer the following translation:
176

177 179 183

In 1986 George Cardona kindly made the pertinent passage of the MS available to me in the form of a very readable xerographic copy from a microfilm in his possession. The facsimile reproduction of the MbhDms published by BORI (Poona 1980) is legible but not very clear. MS tisthatityasmin. m So MS. MS sarvapadddeso but this is a well-known verse, e.g. Mbh 1:75,13-14, see p. 187 with note 18 181 28 above. ° MS samir (?). MS ekah. m MS arthapratyuktanam. MS avirupdndm. m MS dadhyddisyd ikas. 185 MS so rtha sthdnam.

252

Indian semantic analysis

And sthdna is artha 'meaning' [according to the analysis]: [something] occurs (tisthati, lit.: 'stands') in it, that is, it is seen [there].186 [Thus, for example, A 2.4.52 aster bhuh can be paraphrased:] in it, that is, in the meaning of as, bhu (bhavatih)1*7 [occurs]. There is to be188 a cancellation of the other [element, namely as, in the context of an ardhadhdtuka suffix], since (yasmdt) bhu (bhavatih)m [in that context] expresses the meaning of it (=as).m Now, concerning substitution of single sounds, what is the sense of sthdna [then], in as much as meaning does indeed not exist for single sounds? Even in this case [the interpretation as artha] is acceptable, on the view that 'all [substitutes] are substitutes of complete words',191 so that for the [whole] element dadhi there is the substitute dadhy, likewise rup for ruh.191 Alternatively, kdla 'time' is expressed by the word sthdna, so that [a rule such as A 6.1.77 iko yan aci teaches that sounds denoted by] yaN occupy the time of [pronunciation of sounds denoted by] iK. Or, it is so that this as (asteh) [of A 2.4.52] is an imitation [in grammar] of all the units as [of the real language] which are used to signify the meaning [which is to be conveyed] (arthaprayuktdndm). These [units as] are the substituends, and their meaning is provided for193 by bhu (bhavatind). Similarly also in the^ case of [A 6.1.77] iko yan act [the sounds denoted by] iK, those which are not accomplished, [that is, the sounds denoted by] iK which occur in dadhi and such, that which is their own form (svarupam), that meaning (artha) is the sthdna.
186

This would be the translation if the reading drsta of the MS is retained. It is, however, tempting to emend drsta to sabda as was first suggested by Abhyankar and Limaye and retained by Palsule and Bhagavat. Such a wording of the analysis is known from Kaiyata's PradTpa (MbhP I:408b): tisthanty asmin sabdd iti sthdnam, arthe ca sabdas tisthati', cf. p. 256 below. Kaiyata's reliance on the Dlpika and on Bhartrhari in general is well known. 187 The MS has bhavati, but cf.0below. 188 The reading bhdvinivrttis of the MS is thus accepted and considered a compound. Compounds with bhdvin- as their first member are not numerous according to standard Sanskrit lexica; MW gives only bhdvicakravartin 'a future king'. However, the" word bhdvin'future; which is to be' is used several times by Bhartrhari in his VdkyapadTya (e.g. VP 3.9 40, 95, 114). It is noteworthy that Punyaraja, in the context of VP 2.184ab: dhdtoh sddhanayogasya bhdvinah prakramdd yathd, expresses himself by such compounds as bhdvisddhanasambandhasamdsrayanena and bhdvisarfftiandhdt (ed. Subramania Iyer 1983:78). 189 Note that the MS actually reads bhavatih (bhavatis tasartham) here, so that at this point it was Abhyankar and Limaye who emended to bhavati. Palsule and Bhagavat suggest the reading °bhavatih bhavati I nivrttis0. 190 Imagining that Abhyankar and Limaye were right in emending bhavatih to bhavati, one may, according to a rather peculiar syntax which is avoided if the reading of the MS is accepted, translate: 'when there is to be a formation [with a linguistic element] different from the other [element], it signifies the meaning of the latter'. In that case it would be reasonable to accept also the emendation ofbhdvinivrttis0 to bhdvinivrttis0 suggested by Abhyankar and Limaye. Accepting this emendation but retaining the reading bhavatih of the MS, one may also translate: 'there is to be an occurrence of another [element, namely bhu (bhavatih)], since bhu expresses the meaning of it (as)'. 191 This alludes to a verse quoted Mbh 1:73,13-14; see p. 187 with note 28 above/ 192 A 6.1.77 iko yan aci provides for semivowel substitution, in this case y for i, so that dadhi is replaced by dadhy when a vowel follows. A 7.3.43 ruhah po 'nyatarasydjn teaches the optional replacement of rup for ruh when forming the causative stem. 193 The MS reading vidhiyate 'is taught; is provided for' is retained also by Abhyankar and Limaye. The slightly better sense 'and their meaning is denoted by bhu9 may be obtained by emending vidhiyate to (a)bhidhiyate as suggested by Palsule and Bhagavat. -

Substitution

253

This is what Bhartrhari tells us concerning the interpretation of the word sthdna. The ideas put forward in the passage may be paraphrased as follows. First his major thesis is propounded, namely that sthdna refers to the meaning of a linguistic element, as shown by his analysis as adhikaranasddhana and the paraphrase of rule A 2.4.52 aster bhuh. According to this interpretation, the sthdna is the location in grammatical space occupied by a linguistic element. This location is identified as the meaning {artha) in which an element occurs. Since perennial linguistic elements, such as as and bhu, are proper to particular different contexts, one element may be said to occur in the location of another element, that is to say, in the meaning of another element, according to a semantic topography where the two elements are in complementary distribution. Next follows the major objection concerning substitution of single sounds, in which case no meaning is involved. This objection is overcome in three different manners: First, by the principle that a grammatical substitution operation applies only to whole linguistic items, not to single sounds, so that a rule such as A 6.1.77 iko yan aci, which provides for prevocalic semivowel substitution, actually provides for whole elements such as dadhy to occur in domains different from those of other whole elements such as dadhi. Under these conditions substitution involves whole elements which are used for each other in certain locations so that they are in complementary distribution in one meaning within the linguistic practice Paninlyas set out to describe. Alternatively, and in this case giving up the position that sthdna refers to artha 'meaning', Bhartrhari suggests that sthdna is considered as referring to kdla 'time', so that a rule such as A 6.1.77 iko yan aci would provide that a semivowel occupies (adhitisthati) the time of a vowel when another vowel follows. That is to say, at the time for pronouncing a vowel, a semivowel resides there instead. This kdla alternative I shall return to below. Finally, Bhartrhari proposes another possibility, reverting to the view that sthdna signifies artha 'meaning', but now allowing the own form (svarupa) of a linguistic element to be one of the meanings involved. This alternative is introduced in the text by yathd vd, correlated with evam which follows. As I take it, the parallel is this: as (asteh) of A 2.4.52 aster bhuh is a self-referring term whose meaning is all the elements as (asirupdndm) that occur in actual forms, so that they are prompted by a meaning that is to be conveyed (arthaprayuktdndm). And these are the sthdnins, the substituends. In other words, the substituends referred to by as in A 2.4.52 are instances of as actually used to signify meaning. And their meaning is provided for by bhu. The situation is similar in the case of a rule such as A 6.1.77 iko yan.aci: just like as of the rule aster bhuh refers to as as actually used in meaningful forms, in which as has meaning, so the pratydhdra term iK refers ultimately to the vowels / etc. which do have meaning by referring to the actually used sounds i etc. of dadhi and such. And this meaning is the sthdna, the meaning which is the own form of the vowels i etc. ultimately signified.

254

Indian semantic analysis

K

-

Concerning this last line of the DTpikd passage, I am not completely confident of the text or my translation, although I think it is pretty Clear what Bhartrhari has in mind. The MbhDms here reads iko ye na krtd dadhyadisya0. Abhyankar and Limaye have emended this to iko yani krte dadh^ddih sydf which, although not without a certain effort, could be interpreted: 'when yaN has been substituted for iK, dadhi etc. should come about'. Palsule and Bhagavat (1991:19,25-6) suggest iko ye 'nukrtd dadhyddisthd ikas tesdm yat svarupam so 'rthah sthdnam iti which they translate (ibid., trans, p. 71): 'the ik-s in (the real) words dadhi etc. which are named imitatively, - their meaning, viz. the form, is denoted by sthdna (in this rule)'. However, the reading I have proposed above would be more faithful to the actual manuscript reading, emending only dadhyadisya to dadhyddisthd(h) 'which occur in dadhi and such'. I have not been able to make anything out of the possibility of reading yena and not ye na. *\ It is, at least, quite clear from this passage that Bhartrhari propoUpds as his major thesis that sthdna refers to artha 'meaning'. The objections to such a view are carefully considered, and the position is ultimately retained. The alternative interpretation of sthdna as kdla 'time' which is resorted to in order to overcome the difficulty created by substitution of single sounds, I shall return to briefly. At first sight the interpretation as kdla seems to refer to another locus according to an interpretation of sthdna as adhikaranasddhana. But note that the interpretation of sthdna as prasanga 'possible appearance', or any other familiar alternative based upon an interpretation of sthdna as bhdvasddhana, is not discussed in the DTpikd passage at all. Elsewhere in the DTpikd194 Bhartrhari seems to bring up the idea of kdla as an alternative to prasanga. Basic to Bhartrhari's ideas is the notion of sabdabrahman, ultimately speaking the sole reality of undifferentiated being qua Speebh.195 This brahman is one, but it appears as many196 through the power ofJkdla 'time' which is an aspect of brahman.197 Duejxi this kdlasakti 'Power of Time' the world appears through the six modifications of being, sad bhdvavikdrdh, referred to already in the Nirukta (Nir 1.2) and attributed by Yaska to a certain Varsyayani. A detailed analysis of kdla 'time' is offered by Bhartrhari in kdnda 3.9 of his VdkyapadTya.m Although it is ultimately one, Speech appears in sequences due to the Power of Time. Thus it is not surprising that Bhartrhari in his DTpikd identifies sthdna as kdla, that is to say, as that portion of time which is occupied by a linguistic unit at the time of its pronunciation. The idea of Time and Speech is addressed briefly in the VdkyapadTya (VP 1.49-50): nddasya kramajdtatvdn na purvo na paras ca sah I akramah kramarupena bhedavdn ivajdyatell
194 198

For example, MbhD 302,21 f., when discussing vt. 3 siddham tu prasange raparatvdt on A m 197 1.1.51 ur an raparah. 195 VP 1.1. VP 1.2. VP 1.3. Translated with Heiaraja's commentary by P. Sarveswara Sharma 1972.

Substitution pratibimbam yathdnyatra sthitam toyakriydvasdt I tatpravrttim ivdnveti sa dharmah sphotanddayoh 11

255

Since sound is produced in [time-]sequence, this [sphota 'indivisible unit of Speech'], which [can be characterised] neither as earlier nor as later, [and as such is] without sequence, appears as if possessing parts through [sound which has] the form of sequence.199 Just as a reflection [of the moon etc.] which resides elsewhere [than in itself, namely in water,] seems, as a result of movement in water, to follow the activity of this [water] - [such] is that which characterises [the relation between] the indivisible Unit of Speech {sphota) and sound. The Vrtti states in this context (ed. Subramania Iyer 1966:106): °nddah sapratibandhdbhyanujnayd vrttyd sphotam avadyotayati, '. . . through its function which comprises prevention and permission, sound suggests the sphota, the indivisible Unit of Speech'. While bhedavat 'possessing parts' pertains to dravya 'substance', krama 'sequence' pertains to time. The notion of pratibandha 'prevention' and abhyanujfid 'permission' is a function through which time regulates the world.200 With regard to language, this works in such a way that the eternal and indivisible Speech, although one, appears as if it had parts through sound which is manifest in time and space. And through these functions of time, the manifested sounds appear in sequence in such a way that when one is permitted to appear, all others are prevented. In this way the notion of time could be said to fill the function of prasanga 'possible appearance'. An analysis of sthdna as bhdvasddhana would still hold, since time is one and in principle a continuum which is ultimately without sequence, like Speech itself. Kaiyata on sthana Kaiyata deals with both of the common interpretations of the term sthdna, as a noun denoting a locus in the sense of artha 'meaning' and as an action noun in the sense of prasanga 'possible appearance'. He proceeds to show how each of them would affect the analysis of A 2.4.52 aster bhuh (MbhP I:408b):
199

200

This is the interpretation I suggest if the text is read as above. The Benares edition (Patavardhana et al. 1884-1937) and the edition of Sltaramacarl (1926/7) read grhyate for jayate, listed as v.l. in the edition of Abhyankar and Limaye (1965) but surprisingly not in the edition of Rau (1977). The commentary of the Benares edition gives the v.l. sakramah for akramah. Rau (1977) offers the v.l. sakramam kra°. Reading sakramah and grhyate one may translate: This [sphota], which [can be characterised] neither as earlier nor as later, is linked to [the notion of time-]sequence, because sound arises in [time-]sequence. Through [this] form of [time-]sequence [that pertains to sound,] [sphota] is perceived as if possessing parts. By way of comparison, J.F. Staal (1967:vii) translates: Though sound is produced in a fixed order Speech itself has no earlier or later Itself without order, it is produced As if divided by the appearance of order. Cf. VP 3.9.4.

256

Indian semantic analysis

sthdnam artha iti ke cid dhuh tisthanty asmin sabdd iti sthdnam, arthe ca sabdas tisthati I tendster arthe bhur ity arthah II anye tv dhuh bhrasjo ropadhayo ram anyatarasydm ityddau ropadhddindm dnarthakydd bhdvasddhanah prasahgavdcT sthdnasabda dsnyatel tendsteh prasange - asteh prdptau bhuh prayujyate ity arthah sampadyate II svdbhdviko 'ster aprayoga drdhadhdtuke bhusabdasya ca prayogo 'nena prakdrendnvdkhydyate II Some say that sthdna [signifies] artha 'meaning' [according to the analysis]: 'something occurs in it, namely linguistic elements, thus [it is called] sthdna\ and a linguistic element occurs in [its] meaning. Hence the sense [with reference to rule 2.4.52 aster bhuh] is that 'bhu [occurs] in the meaning of as\ But others say that, since in [rules] such as [A 6.4.47] bhrasjo ropadhayo ram anyatarasydm the sound r and the penultimate etc. are meaningless, the word sthdna is formed with the sense of the mere activity [of standing] and signifies prasanga 'possible appearance'. Therefore the sense brought forward is that when there is a possible appearance of as, that is, when as would [otherwise] obtain, bhu is used. The non-use of as which already exists (svdbhdvika) in the domain of an dr\dhadhdtuka suffix and also the use of the element bhu are explained in this way. Kaiyata first notes the interpretation of sthdna as a derivative signifying a locus, in particular the meaning (artha) in which a linguistic element occurs. In this connection, recall that expressions like ayam sabdo }sminn arthe vartate, comparable to English This word occurs in this meaning', are normal Sanskrit and that Paninlyas refer to the locative in such usages as a visayasaptamT, a locative which denotes a domain.201 Thus, for example, rule 2.4.52 aster bhuh together with rule 1.1.49 would, under this interpretation, provide that bhu occurs in the meaning of as (aster arthe) in the domain of drdhadhdtuka suffixes. That is to say, bhu is used in the domain of drdhadhdtuka suffixes to signify the meaning which as signifies in the domain of sdrvadhdtuka suffixes. This implies that linguistic elements 'stand' in their location in semantic space.
201

For example, Mbh 11:199,21 (Bhdsya on A 4rtt stnydm, vt 5): stnydm yafprdtipadikam vartate tasmdt tdbddayo bhavanti, 'when a nominal stem occurs in the domain of the feminine, [the suffixes] TaP etc. apply'; similarly Mbh 11:31,10 (on A 3.1.26): bhavantiha hi visayasaptamyo 'pi I tad yathdl pramdne yat prdtipadikam vartate stnydm yat prdtipad vartata iti, 'for here also locatives which denote a domain apply: for example, "when a nominal stem occurs in the domain of a measure; when a nominal stem occurs in the domain of the feminine"'. See also Cardona 1983:44, note 80. Another instance, which involves rule 2.4.52 aster bhuh, is A 2.4.35 drdhadhdtuke. The Kdsikd remarks (Kas 11:284): visayasap tami ceyam na parasaptamil tendrdhadhdtukavivaksdydm ddesesu krtesu pascdd ya prdptam pratyayd bhavanh, 'and this is a locative which denotes a domain, hot a locative which denotes that which follows; therefore, when one intends to use an drdhadhdtuka suffix, then, after the ddesas have been made, the suffixes are added accordingly'. The problem here is that there are two rules which introduce two different gerundive ikrtya) suffices, A 3.1.97 aco yat and A 3.1.124 rhalor nyat. A 3.1,97 teaches that after a root which ends in a vowel, you add yat, which, applied to ybhu as the ddesa of ^as in the domain of drdhadhdtuka suffixes, correctly yields bhu-ya —»bho-ya - - bhav-ya. A 3.1.124 teaches that nyat is added to a «> root which ends in r or a consonant. Now ^as ends in a consonant. If we interpret drdhadhdtuke as a parasaptami one would add nyat to as and eventually end up with the erroneous form bhdvya; so it is considered a visayasaptami. See also Mbh 1:484,5-8.

Substitution

257

Under the interpretation as bhdvasddhana, on the other hand, the term sthana signifies prasanga 'possible appearance' and the same rule 2.4.52 would provide that bhu occurs in the domain of ardhadhdtuka suffixes where as would otherwise obtain. As Kaiyata notes, some insist that only this interpretation is justified, since in some rules for the interpretation of which A 1.1.49 must come into play the sthdnin is not a meaningful element. Single sounds, to which meaning is not normally ascribed by the grammarians,202 are also signalled to undergo substitution operations by rule 1.1.49, and in such instances it is difficult to give the sense of 'meaning' to the word sthana. For example, A 6.4.47 bhrasjo ropadhayo ram anyatarasydm should provide for optionally not using the r of bhr- and the penultimate s of -sj in ^bhrasj 'fry' in the domain of an ardhadhdtuka suffix, adding instead the augment rAM (r) after the vowel. In short, the rule provides for using bharj instead of bhrasj. But r and s are not meaningful elements. This rule presents us with interpretational difficulties on its own.203 Specifically, rAM applies in the place of the r and the penultimate, that is, s, while A 1.1.47 mid aco yntydt parah teaches that an element marked with an anubandha M, such as rAM, is added after the final vowel of the form to which it applies. So the M directs the augment to one place, that is, after the a of bhrasj, but the sthana is stated to be in the place of the r and the s. Accordingly, one is left with a sthana that is at another place. This could be said to be held by lopa 'elision' or zero, since aprayoga 'nonuse' is adarsana 'non-seeing', and adarsana is lopa.204 This is merely a line of reasoning; the texts talk only of nivrtti 'cancellation' by the sthdnasasthT. The Kdsikd considers the case of A 6.4.47 bhrasjo ropadhayo ram anyatarasydm as follows (Kas V:398): ropadhayoh iti sthdnasasthlnirdesdd upadhd rep has ca nivarttete / mittvdc cayam aco 'ntydt paro bhavatil Because a genitive which marks the sthana is exhibited in the word ropadhayoh, the penultimate and the r are cancelled. But since it is marked with the diacritic marker M, this [r] appears after the final of the vowels [according to A 1.1.47 mid aco 'ntydt parah]. On this view, the disappearance of r and s is signalled by the genitive case suffix, while the position of r in forms affected by rule 6.4.47 bhrasjo ropadhayo ram anyatarasydm is said to be signalled by the anubandha M of rAM as stipulated by rule 1.1.47 mid aco yntydt parah. According to the Kdsikd, then, it is the disappearance of s and r in words formed by the addition of ardhadhdtuka suffixes, such as the infinitive bharstum 'to fry', which is due to the genitive case, while the position of r in such forms is effected through the anubandha M of rAM. The locus classicus for the discussion of non-meaningful elements is the Mahabhdsya at sivasMra 5, vt. 9 ff. (Mbh 1:30 ff.). 203 Cf. Patanjali's discussion at A 6.4.1 (Mbh 111:199,17 f.); cf. Benson 1990:95-7. 204 By A 1.1.60 adarsanam lopah.
202

258

Indian semantic analysis

Be this as it may, an infinitive form bharstum will optionally obtain by adding the drdhadhdtuka suffix tumUN {-turn) and by replacing thej of bhrasj by s by A 8.2.36 vrascabhrasjasrjamrjayajardjabhrdjacchasam sah. Otherwise one will get bhrastum. What is important here is that neither the r nor the s possess,meaning on their own. Still their replacement is signalled by a sthdnasasthi. This makes it difficult to say that rule 6.4.47 bhrasjo ropadhayo ram anyatarasydm teaches forms which are used in the meaning of r and s. Kaiyata attributes this view to others. This may indicate that it is not his own position, but it is perfectly possible that he merely presents two poitats of view without any particular preference for one or the other. But he musi have been familiar with arguments that rAM in A 6.4.47 is not an irrefutable objection against the position that sthdna means artha since Bhartrhari, as became evident earlier on, defends this position in his DTpikd, a work with which Kdiyata was not only familiar, but which he draws upon extensively in his own Prddtpa. The matter is also discussed by Nagesa in a passage that will be dealt with shortly. The meaning of the term sthdna is in this case identified as prasanga 'possible appearance', so that A 6.4.47 could be said to teach forms where, under given conditions, rAM applies when the possible appearance of ir and s would otherwise obtain. Kaiyata argues that this is possible on the view that grammar only deals with words that already exist, that are svdbhdvika 'naturally existent'. The non-use of as and the use of bhu as taught by A 2.4.52 aster bhuh are explained in this way as well, the two linguistic elements thus having complementary domains. That sthdna signifies prasanga means that technically, in grammar, one would hear as if it were not cancelled by the body of rules. Similarly, with regard to A 6.4.47, one would hear the r arid the s if they were not cancelled. I do not think Kaiyata is right here, for prasqkta forms are just those which the rules have led you to so far, not forms which are svdbhdvika 'naturally existent'. Nagesa on sthana <

In his Uddyota Nagesa comments on the last-cited PradTpa passage in a way that sheds considerable light on the interpretation of sthdna as artha 'meaning'. He remarks on the two interpretations of sthdna as follows (MbhU I:408b-9a): * arthe cetil vdcakatvasambandheneti bhdvahll asteh prasanga ity asya vydkhyd asteh prdptdv itill prayoga hi I svdbhdvika ity anusajyatel sabdan\tyatvabhangapattyd sthdnasabdo nivrttivdcT nopapadyata iti bhdvahl ramvidhau ropadhaghatitasyarthe ramghatita ity arthe na ksatir ity atrarucihll 'And [a linguistic element occurs] in [its] meaning.' Through the relation of being [its] signifier, that is the idea. The explanation of 'when there is possible appearance of as' is 'when as would [otherwise] obtain'. [On] 'use': [The word] svd-

Substitution

259

bhdvika 'naturally existent' adheres [to this as well, and not only to aprayoga 'nonuse'].205 The idea is that it is not acceptable that the word sthdna has the sense of nivrtti 'cancellation; ceasing' since this would incur destruction of [the view that] linguistic elements are eternally established. [But in fact] there is no harm [to the view that sthdna means artha] if the sense conveyed is that when rAM is taught [in rule 6.4.47], an item which has rAM as a constituent [occurs] in the meamng of one that has r and the penultimate [s] as constituents - and therefore all of this is unsatisfactory.206 Here Nagesa makes it clear that a word - or, more precisely, a linguistic element - may be said to 'occur in' its meaning through the relation of being its signifier. The prasanga alternative is then discussed. With regard to this he brings in the notion of sabdanityatva, the position that speech forms are eternal, that linguistic elements are permanently established. This, according to Nagesa, is what Kaiyata has alluded to by the word svabhavika 'naturally existent'. All that is explained by grammar is the use of bhu and the non-use of as in the domain of drdhadhdtuka suffixes. The elements bhu and as have of their nature complementary domains. One has not established the use of bhu and the non-use of as. Words exist already. Taking sthdna in the sense of nivrtti 'cancellation; ceasing' is thus in principle incompatible with the view that speech forms are eternal, which is also incompatible with interpreting sthdna in the sense of apakarsa 'drawing away'. This view is expressed also in the PadamanjarT on the Kdsikd at A 1.1.49 (KasP 1:169; similarly the Nyasa, KasN 1:168): mvrttivdcT na grhyate, aster upadesasdmarthydt I na hy drdhadhdtuke nivarttitasya sdrvadhdtuke sravanam upapadyate, na hi mathurdydm vydpdditah srughne jfvati I ndpy apakarsavacanah, na hy arthena nityasambaddhasya tato 'pakarsah sambhavatil ndpy drdhadhdtukdd apakarsah, na hi nitye prakrtipratyayasamuddye prakrter apakarsah sambhavati I svabhavika eva hy aster aprayoga drdhadhdtuke bhusabdasya ca prayogo 'nena prakdrendnvdkhydyate, atah prasangah eva sthdnaml [That sthdna] is expressive of nivrtti 'ceasing' is not understood by force of the teaching of as [in the grammar]. For it is not appropriate that something which has ceased [to exist] before an drdhadhdtuka suffix should be heard before a sdrvadhdtuka suffix. Someone who has been killed in Mathura is not alive in Srughna! Also
205 206

The term svabhavika qualifies both prayoga 'use' and aprayoga 'non-use' in the PradTpa passage Nagesa comments upon That is to say, that A 6 4 47 should make one prefer the prasanga alternative is not a compelling argument Although I do not think so, it is theoretically possible that the word arucih 'dislike' in this passage refers to Kaiyata's dislike. If so, the last sentence must be translated '[Kaiyata has expressed] dislike here because there is no harm [to the position that sthdna should be interpreted as artha] if the sense conveyed is that when rAM is taught [in rule 6 4.47], an item which has rAM as a constituent [occurs] in the meaning of one that has r and the penultimate [s] as constituents'. Under this interpretation, Kaiyata too would defend the position that sthdna means artha

260

Indian semantic analysis

i [it does] not [obtain that sthand] expresses apakarsa 'drawing away'. For neither is drawing away of something which is eternally related to [its] meaning possible, nor is drawing away from an ardhadhdtuka suffix, since, in as much as the aggregate of stem and suffix is eternally established, drawing away of the stem is not possible. For the naturally existent non-use of as in the domain of an ardhadhdtuka suffix and also the use of bhu are explained in this way. Accordingly, only prasanga 'possible appearance' is [the sense of] sthdna. Haradatta thus rejects the idea that sthdna should mean nivrtti 'ceasing; cancellation'. Something which has completely ceased to exist in one place does not appear at another. In other words, if an element such as as were cancelled, if it were cancelled forever, there would be no point in mentioning as again at all since it would have been destroyed for good. Still, a substitution process involving nivrtti could be justified by saying that what changes is actually a cognition onaJmowledge^(fewJJ/z/, sampratyaya), that is to say, perennial elements proper to different contexts are conceptualised. One cognition is replaced by another cognition in the mind. By saying that the replacements actually take place in the mind, the doctrine of nityatva is secured even on the view that sthdna has the sense of nivrtti. This view is discussed at various places in vydkarana literature, for example, in the Mahdbhdsya on A 1.1.56, vt. 14 kdryaviparindmdd vd siddham, 'or [the doctrine of nityatva is accomplished] because there is a transformation of what is going to be produced'. Patanjali takes the word kdrya- here as a feminine adjective denoting buddhi 'mind; cognition', arguing that only buddhi undergoes a viparindma 'transformation' (Mbh 1:137,13-18): atha vd kdryaviparindmdt siddham etatl kim idam kdryaviparindmdd itil kdryd buddhih sd viparinamyate I nanu ca kdryaviparindmdd iti bhavitavyarf^ I santi caiva hy auttarapadikdni hrasvatvdnil api ca buddhih sampratyaya ity anarthdntaram/ kdryd buddhih kdryah sampratyayah kdryasya sampratyayasya Viparindmah kdryaviparindmdd itil parihdrdntaram evedam matvd pathitam katham cedam parihdrdntaram sydtl yadi bhutapurve sthdnasabdo vartatel Or else this [position of nityatva] 'is accomplished because of a kdryaviparindma\ What is this kdryaviparindmdtl A cognition to be produced, that is what undergoes transformation. Well, then [the wording] should have been kdrydvipdrindmdt. But there are instances of shortening which are dependent upon a following word [cf. A 6.3.63]; moreover, buddhi and sampratyaya 'knowledge' do not differ in meaning, [and so you can talk of] a cognition to be produced [as equivalent to] a knowledge to be produced, [and accordingly of] a change in a knowledge to be produced as a transformation-to-be-produced, and accordingly [defend the nityatva position by arguing:] 'because there is a transformation of what is to be produced7. This [vdrttika] is read on the assumption that it in fact presents an alternative solution [to the problem], but how could this be an alternative solution? If the word sthdna is used for something which was there previously. Nagesa remarks on the last statement (MbhU I:454a):

Substitution

261

bhutapiirve sthdnasabdaprayogam abhyupetyedam parihdrdntaram I anyathdnarthakyaprasangdt II na ca tatra sthdnasabdo drsyate, somasthdne itydddv abhutapiirve eva prayogadarsandd iti yadi bhutapiirve ityddi bhdsyatdtparyam I This is an alternative solution when it is admitted that the word sthdna is used for something previously existing, because it would not make sense otherwise; but the word sthdna is not seen to possess that sense (tatra - bhutapurva), that is the sense of the Bhdsya on yadi bhutapiirve, because one sees its use only for something which is not previously existing in phrases such as 'in the place of Soma'. That is to say, Nagesa makes it clear that the word sthdna is not used for bhutapurva 'previously existent'. Patanjali continues (ibid., 18-26): bhutapiirve cdpi sthdnasabdo vartate I katham I buddhydl tadyathdl kas cit kasmai cid upadisati prdcinam grdmdd dmrd iti I tasya sarvatrdmrabuddhih prasaktdl tatah pascdd aha ye ksirino 'varohavantah prthuparnds te nyagrodhd iti I sa tatrdmrabuddhyd nyagrodhabuddhim pratipadyate I sa tatah pasyati buddhydmrdms cdpakrsyamdndn nyagrodhdms cddhvyamdndn I nityd eva ca svasmin visaya dmrd nityds ca nyagrodhd buddhis tv asya viparinamyate I evam ihdpy astir asmd avisesenopadistah I tasya sarvatrdstibuddhih prasaktdl so 'ster bhur bhavatity astibuddhyd bhavatibuddhim pratipadyate/ sa tatah pasyati buddhydstim cdpakrsyamdnam bhavatim cddhfyamdnam I nitya eva ca svasmin visaye 'stir nityo bhavatir buddhis tv asya viparinamyate II And the word sthdna does occur with regard to [something] which was there previously as well. How? By [referring to] a cognition. Just as someone points out to someone else: 'there are mango trees east of the village'. That [person] will have the cognition of mango trees everywhere [east of the village]. Then later he (the speaker) says: 'those [trees] with a milky juice, which grow downward, with broad leaves, they are banyan trees'. Then he (the listener) obtains a cognition of banyan trees through the cognition of mango trees. Then he sees that the mango trees are removed and the banyan trees are established in his mind, but the mango trees are in fact permanently established in their own domain, and the banyan trees are permanently established [in theirs]. Only his cognition is transformed. Similarly also in this case: [the verbal root] as has been taught to him without specification, so he will have a cognition of as everywhere, [that is, in the domain of sdrvadhdtuka as well as that of drdhadhdtuka suffixes]. He then obtains the cognition of [the verbal root] bhu through the cognition of as by [A 2.4.52 which teaches that] bhu appears in the place of as [in the domain of drdhadhdtuka suffixes]. Then he sees that as is removed and bhu is established in his mind, but as is in fact permanently established in its own domain, and bhu is permanently established [in its]. Only his cognition is transformed. Kaiyata remarks (MbhP I:454a): na ca nityatvavirodhah I buddhidharmdh kevalam arthesu pratipattrvasdd dropyante, na tu tadvasdd arthdndm tathdvasthdnam, 'and so there is no violation of the nityatva view: it is only the properties of cognitions that are imposed upon things by force of the cogniser; but it is not by such power that the condition of things is [established] as such'. Nagesa comments on this as follows (MbhU I:454b): na ceri/ bhutapiirve hi

262

Indian semantic analysis

sthdnasabdo nivrttivacanah sydd iti bhdvahll tatpakse nityatvam katham id prcchati bhdsye katham iti, '"and so there is no": for if the word sthdna were used for something which was previously, it would express cessation {nivrtti), that is the idea; and on that view, how could there be nityatva, and so he asks, in the Bhasya, "how?"'. The sense of nivrtti is thus retained on the view that what is replaced in a substitution process is actually a cognition* while the linguistic elements themselves are perennially established in their respective domains.207 Haradatta also rejects the sense of apakarsa 'drawing away', on the basic premise of Paninlyas that the relation between a linguistic element and its meaning is eternally established, not arbitrary or invented by someone. This is a position held also by the Jain grammarian Abhayanandin, .author of the Jainendramahdvrtti on Devanandin's Jainendravydkarana (JV 1.1.46): nityasabddrthasajnbandh'avivaksdydm sthdnasabdah prasangavdcTI prasahgas ca prdptdrhatvam svdrthapratydyakdvasaro vdl yathd guroh sthdne sisya upacaryate iti guroh prasahga iti gamyate I evam asteh sthdne prasange bhuhl bhavitdl bhavituml bhavitavyam I brunah prasange vacir bhavatil vaktdl vaktuml vaktavyaml anityasabddrthasambandhavivaksdydm apakarsavdcT sthdnasabdah I yathd goh sthdne asvam badhndtil evam asteh sthdne 'pakarse bhur bhavatil When one intends to express [the position] that the relation between linguistic elements and their meanings is eternally established, [then] the word sthdna expresses prasahga 'possible appearance'. And prasanga [is either] the condition that entitlement obtains208 or [it is] an occasion which expresses the sense of the thing [in question]. For example, The pupil is served in the place of the teacher' means 'when there is entitlement for the teacher'. Similarly, in the place of as, that is, when the occasion arises [for as], bhu [applies]. As in [the following formations with drdhadhdtuka suffixes]: bhavitd, bhavitum, bhavitavyam. When the occasion arises for bru, vac applies. When one intends to express [the position] that the relation between linguistic elements and their meanings is not eternally established, [then] the word sthdna expresses apakarsa 'drawing away'. Just as [when one says] 'He ties up a horse in the place of theTcow.' Similarly, in the place of as, that is, when there is drawing away [of as], bhu applies. Thus, the sense of apakarsa is rejected on the premise that the sense of prasanga obtains because the relation between a linguistic element and its meaning is eternally established. Logically, this is of course not the only circumstance under which prasanga may obtain. The implication of the view expressed by Abhayanandin is that the two alternatives, prasanga and artha, overlap. Panini's grammar is a system where derivations start from semantics. When prasanga obtains it is because the relation between a linguistic unit and the meaning in which it occurs is considered to be eternally established.
207 208

The term buddhivipanndma 'transformation of cognition' and the problem of illusory change have been discussed by D.S. Ruegg 1958. Or, possibly: 'the condition of deserving [to receive or undergo] what has arrived [i.e., is at hand]'.

Substitution

263

Similarly a stem and a suffix are semantically connected within a system of mutually excluding domains. What then about linguistic items which are not considered meaningful elements but which are nevertheless affected by substitution operations? The Nagesa passage, to which I shall now return, could be interpreted so as to offer a possible solution to this problem. Rule A 6.4.47 bhrasjo wpadhayo ram anyatarasydm teaches an option whereby the r and s of the verbal root Abhrasj are not used but the augment rAM (r) is inserted after the vowel of the root. But single sounds such as r and s are not considered meaningful elements, and so sthdna cannot mean ariha. Nagesa dismisses this view and claims that there is no harm to the position Kaiyata originally presents - namely that sthdne is interpreted as arthe - on the assumption that A 6.4.47 provides for using an item that has rAM as a constituent in the meaning of one that has r and penultimate s as constituents (ropadhaghatitasyd). I think that this is what Nagesa has intended with the claim that 'there is no harm' (aksatih). Another possibility would be that aksatih refers to the absence of harm to the nityatva position. But this is hardly tenable, since it would be true also under the interpretation that sthdna means prasanga. Rather, once one interprets the rule as shown, there is no harm done to the view that sthdna has reference to meaning. Such an interpretation is supported also by what Nagesa has stated elsewhere, a circumstance that weakens the possibility that the word aruci 'dislike' expresses Kaiyata's view. Rather, the dislike is Nagesa's own, something which is supported by what he says in his Laghusabdendusekhara. There he takes the term sthdna of A 1.1.49 in the sense of prasanga 'possible appearance' but remarks that, insofar as possible, the prasanga in question is of a meaningful item. To illustrate this point he uses A 6.1.77 iko yan aci which thus provides for the use of an item with a prevocalic constituent y to convey the meaning of an item with a prevocalic constituent / which would otherwise possibly appear to convey the meaning (LSI 58,6-9,2): prasahgas ca sati sambhave 'rthavata eva, arthapratydyandrtham sabdaprayogdtI iko yan itydder acparekdraghatitasabdasydrthabodhandya prasange tatparayakdraghatitah prayoktavya ity arthahl And possible appearance, insofar as possible, is only of something that is meaningful, since the purpose of linguistic elements is to bring about understanding of meaning. That is to say, according to [A 6.1.77] iko yan etc., when the occasion arises {prasange) for communicating the sense of a linguistic element which has / as a prevocalic constituent, [an element] with a prevocalic constituent y is to be used [instead]. The possibility Nagesa has brought up is, of course, compatible with the position that one considers whole items such as dadhy to appear in certain domains instead of items such as dadhi, which of their nature have complementary domains. The Sanskrit grammarians who believed in the doctrine of

264

Indian semantic analysis

nityatva were forced to adopt a model of substitution where substitutes are considered substitutes of whole words. This position is mentioned already in a verse quoted in the Mahdbhdsya (Mbh 1:75,13—14).209 Nagesa's position is also compatible with the idea that what actually changes is a cognitiorif210 Thus the Laghusabdendusekhara passage continues (LSI 59,2-6): er uh ityades ca tes tuh iti; ato na sabdanityatdhdnih I kim ca ikdrabuddhiprasange yakdrabuddhir ity arthahl ata eva acah parasmin ityddi sangacchatel spastas cedam dddhd ghu iti sutrasthdnivatsutrayor bhdsyel And according to [rules] such as [A 3.4.86] er uh}xx there is -tu for -ti,m and so the position that linguistic elements are eternally established is not forsaken. Moreover, there is a cognition of y when there is the possible appearance of a cognition of i, that is the meaning. For this very reason [A 1.1.57] acah parasmin etc.213 suits. And this is clear from the Bhdsyam on rjules [A 1.1.20] dddhd ghv [addp] and [A 1.1.56 which teaches that the ddesa is considered] like the sthdnin. On the view that linguistic elements are naturally existent (svdbhdvika), as is eternally established but does not appear before an drdhadhdtuka suffix where the eternally established unit bhu appears. So sthdna cannot mean nivrtti: there is no cancellation, no ceasing. The situation is simply that one element does not appear, since another element appears to convey the meaning according to a system where the two elements have complementary domains. In the same way one can consider a rule such as A 6.1.77 iko yan aci. Once whole units such as dadhy are considered to appear in certain donjains instead of units such as dadhi, it becomes possible to claim that the rule teaches the use of a unit that has y as a constituent in the meaning of one that has i as a constituent, provided a vowel follows. Nagesa is in agreement with Bhartrhari, then, when he brings forward the argument that substitutions are of whole units. Looked upon in this way, it is whole units such as dadhy and dadhi, having of their nature complementary domains, which are the elements mvolved in a substitution process. One whole unit appears in the meaning of another whole unit which would possibly appear to convey the meaning. Therefore you cannot base an argument that sthdna means prasanga and not artha on the view that individual sounds are meaningless. Accordingly, rAM in A 6.4.47 bhrasjo ropadhayo ram anyatarasydm does not provide a critical argument against taking sthdna in the sense of artha 'meaning'.

209 211

210 See p 187 with note 28 above Mbh 1:137,13 f, discussed above Rule 3 4 86 teaches the replacement of i by u in the imperative ending 22 p o r mstance, bhavatu for bhavati. 1 213 Rule 11 57 acah parasmin purvavidhau teaches that the replacement for a vowel is considered like the vowel when there is application of a rule which would affect that which precedes the vowel, and when the replacement is conditioned by what follows the voweL Here the principle of sthdnivadbhdva taught by A 1.1 56 is at work, that is, the principle that the ddesa is considered like the sthdnin. m Mbh 175,8 f, 137,13 f.

Substitution Concluding remarks

265

The results of this inquiry into the meaning of the term sthdna can be summarised as follows. Basically, sthdna is analysed in two different ways, either as bhdvasddhana, with LyuT (-ana with presuffixal accent) by A 3.3.115 lyut ca, that is, as an action noun denoting mere activity, state of being, equivalent to sthiti 4a standing', or as adhikaranasddhana, with LyuT by A 3.3.117 karanddhikaranayos ca, that is, as a noun denoting a locus for the activity denoted by the verbal root ^Isthd. Under the first alternative commentators consider three particular contextual interpretations of sthdna, namely apakarsa 'drawing away', nivrtti 'cancellation; ceasing', and prasanga 'possible appearance; chance'. The particular senses in question are considered contextual varieties of one single general meaning of sthdna. Paninlyas generally take sthdna in sthdneyogd of rule A 1.1.49 in the sense of prasanga 'possible appearance'. In principle, taking sthdna in the sense of nivrtti 'cancellation' or in the sense of apakarsa 'drawing away' is incompatible with the view that speech forms, as well as the relation between a linguistic element and its meaning, are perennial, eternally established (nitya). Still, a substitution process involving the senses of nivrtti or apakarsa could be justified by saying that what changes is actually a cognition. Perennial elements proper to different contexts are conceptualised. The interpretation of sthdna as prasanga 'possible appearance' is clearly applicable in any context where grammatical form alone is considered. This requires a grammatical universe as the range of application, within which one can talk of prasanga or possible appearance of linguistic elements. This universe is provided by the body of rules, on the view that grammar is concerned with linguistic elements that already exist. Grammar does not establish these elements. Thus, with regard to a rule such as A 2.4.52 aster bhuh, all that is explained by grammar is the use of bhu and the non-use of as. The units as and bhu have of their nature complementary domains. Under the interpretation of sthdna as prasanga, a rule such as A 2.4.52 aster bhuh would, together with A 1.1.49, provide that bhu occurs in the domain of drdhadhdtuka suffixes, where there would otherwise be the possible appearance of as. Now, in Panini's grammar derivations start from semantics. The interpretation of sthdna as prasanga obtains because speech units are eternally established, as is the relation between a linguistic unit and the meaning in which it occurs. Thus it is not surprising that commentators also consider an alternative interpretation of sthdna, namely as a derivative signifying a locus, in particular the meaning, artha, in which a linguistic element or speech unit, sabda, occurs. A linguistic unit may be said to 'occur in' its meaning through the relation of being its signifier, as Nagesa notes. One should in this connection keep in mind that an expression such as ayarn sabdo 'sminn arthe vartate, comparable to English This word occurs in this meaning', is normal Sanskrit and that

266

Indian semantic analysis

Paninlyas refer to the loqative in such usages as a visayasaptami, a locative denoting a domain. Thus, under the interpretation of sthdna as artha, rule A 2.4.52 aster bhuh together with rule A 1.1.49 sasthisthdneyogd would provide that bhu occurs in the meaning of as {aster arthe) in the domain of drdhadhdtuka suffixes. That is to say, bhu is used with drdhadhdtuka suffixes to signify the meaning which as signifies in the domain of sdrvadhdtuka suffixes. This implies that linguistic units occur in their respective domains in semantic space, all units and meanings thus embedded in a system where meaning is the locus in which a linguistic unit occurs. One major objection to this interpretation of sthdna as artha 'meaning' concerns rules which seemingly provide for substitution of single sounds, in which case there is no meaning involved. Thus, as Kaiyata notes, some insist that only the interpretation of sthdna as prasanga 'possible appearance' is justified, since there exist rules which are affected by rule A 1.1.49 where the linguistic items stipulated as sthdnin and ddesa are not directly meaningful. However, according to Nagesa, there is no harm to the position that sthdna is interpreted as artha since a rule which provides for substitution of single sounds actually provides for the use of a whole item that has one sound as a constituent in the meaning of a whole item that has another sound as a constituent. The same point is brought forward by Nagesa elsewhere. Thus, in his Laghusabdendusekhara he takes sthdna in the sense of prasanga 'possible appearance' but remarks that, insofar as possible, the prasanga in question should be of a meaningful unit.215 This is compatible with the position referred to already by Patanjali, that in a substitution process one should consider whole items such as dadhy to occur in certain domains instead of other whole items such as dadhi, not only y instead of i. Tn this way the difficulty concerning substitution of single sounds is overcome. Linguistic elements such as dadhi and dadhy are in complementary distribution, or, according to the Sanskrit terminology, such elements are naturally existent (svdbhdvika), having of their nature complementary domains. ' Bhartrhari too alludes to this in his DTpikd, where his major thesis clearly is that sthdna refers to artha, the meaning of a linguistic element. Alternatively, and giving up the position that sthdna refers to artha, Bhartrhari also suggests another possible interpretation of sthdna, namely kdla 'time'. That is to say, sthdna denotes that portion of time which a linguistic item occupies at the moment of its pronunciation. In this way the objection concerning substitution of single sounds is overcome as well. But Bhartrhari mentions lstill another possibility, reverting to the view that sthdna signifies artha, basedon the position that all linguistic items are first self-referring, a position already alluded to by Katyayana (vt. 2 on rule A 1.1.68). Thus, say that the substitudnds referred to by as in rule A 2.4.52 aster bhuh are instances of as actually used to signify
215

When, may one ask, is this not possible7

Substitution

267

meaning, their meaning, then, is expressed by bhil. The situation is similar in the case of a rule such as A 6.1.77 iko yan aci which signals prevocalic semivowel substitution. Now, the pratydhdra terms iK, yaN, and so on are technical terms which signify i and so on, and the vowels ultimately signified also have meanings, namely their own forms (svarupa). Even these forms, then, can be considered to have meanings, namely the sounds i and y that occur in actual terms such as dadhi and dadhy. It is thus clear that the interpretation of sthdna as prasanga 'possible appearance' first and foremost applies at the level of grammatical form. It is equally clear that the alternative interpretation of sthdna as artha 'meaning' is seriously considered or even preferred by three major Paninlyas in the tradition, namely by Bhartrhari in the fifth century, by Kaiyata in the early eleventh century, and by Nagesa in the late seventeenth to eighteenth century. The main objection to the interpretation of sthdna as artha concerns substitution of single sounds where there is seemingly no meaning involved. This objection is overcome in different ways. Both interpretations of sthdna, as prasanga and as artha, presuppose a view of grammar where linguistic units occur in their respective domains according to a system of complementary distribution. This domain may be regarded as basically determined by the possible appearance of a unit or by its meaning. There is a certain choice, then, as to whether one wishes to concentrate on linguistic form or on semantic content, and the two alternatives thus overlap. On the view that speech units are perennial and that the relation between a speech unit and its meaning is eternally established as well, one unit does not cease to exist when a complementary unit appears in its place. The unit which is replaced disappears from the scene, but it does not cease to exist. When the meaning conveyed by as in the domain of sdrvadhdtuka suffixes is conveyed by bhu in the domain of drdhadhdtuka suffixes, as does not cease to exist, it is simply not the placeholder in that particular context.

6 Epilogue

The nirvacana material from the Nirukta makes it clear that in the technical language of Yaska a genitive is used to mark the relation between a term and its explanatory expression. In other words, although an ablative also occurs, there is indisputably a genitive construction to be accounted for. J.S. Speijer (1886:81-2) introduces his description of the syntactic usage of the genitive in the following way: T h e fundamental notion of the genitive or sixth case is to mark the belonging to, partaking of. In Sanskrit, it is employed in, so manifold and so different ways as to make it very difficult to give a satisfactory account of all of them.' This was a fact alsQ to the Sanskrit grammarians. In the words of Patanjali (Mbh 1:118,10-11): ekasatam sasthyarthd ydvanto vd te sarve sasthydm uccdritdydm prdpnuvantj, 'there are one hundred meanings for the genitive case suffix - or as many as there are - and all of them would apply when a genitive case ending is uttered'.1 Now, anyone engaged in nirvacana analysis would be well versed in vydkarana and hence familiar with the use of the sthdnasasthT in the AstddhydyT. Indeed, in Yaska's outline of nirvacanasdstra he explicitly states that the analysis of single words should not be taught to anyone who is not trained in vydkarana.2 On the basis of the previous discussion, however, I think it is possible to put forward a stronger argument here. If, as Kaiyata and others claim, the sthdnasambandha or relation marked by a substitutional genitive is considered antaranga, that is to say, that it comes to mind before any other relation when we encounter a genitive form in the grammar, then there is nothing remarkable in interpreting a genitive as a substitutional genitive in the context of relations between linguistic elements. This is also buttressed by the use of the genitive to indicate synonyms and the usage of the genitive in the formula ity apy asya bhavati met with in the Nirukta. Nor is there anything in the discussions of A 1.1.49 sasthi sthdneypgd and the nirdisyamdna-paribhdsd which prohibits such an interpretation within or without the boundaries of vydkarana. As noted by Cardona (1976:201): It is noteworthy that Panmlyas, from Katyayana on, do not view rules 1.1.49, 1.1.66-67 as stating any syntactic uses different from the object language Sanskrit.
1

See pp. 197 ff above.

2

Nir 2.3, see p. 38 above.

Epilogue

269

Instead, according to them, the metarules in question apply to provide single unambiguous interpretations of rules which, otherwise, the user of the grammar, a speaker of Sanskrit, could not interpret uniquely. Note that I am not claiming that the genitive used to mark the relation between a term and its explanatory expression was intended exclusively as a sthdnasasthi in the technical sense of the Astddhydyl All I am arguing is simply that all genitives are complex and that the substitutional use of the genitive is part of the Sanskrit language so that any genitive in a suitable context could be interpreted in such a way. With regard to the relation sustained between the two terms of a nirvacana analysis, suffice it to repeat Vaidyanatha's remark that on the view that the term sthdna signifies artha 'meaning', the sthdnasambandha is what comes to mind before any other relation because meaning comes first whenvwe use words. Viewed according to a substitutional model, a nirvacana analysis provides two terms, a sthdnin or substituend and an ddesa or substitute, which are related through the sthdna or location in grammatical space which is the necessary condition for the relation in question. When the sthdna is identified as artha 'meaning', the two elements are simply alternative placeholders in semantic space, and artha 'meaning' is the necessary condition for the relation between a word and its explanatory expression. Meaning, it may be recalled, is the primary quest of nirvacana analysis. According to A 1.1.56 sthdnivad ddeso 'nalvidhau the ddesa is treated like the sthdnin except in a rule teaching an operation that depends on sounds. In other words, the characteristics of the substituend are transferred to the substitute. This principle is valid also in the ritual Sutras where ritual elements are substituted for each other or for cosmological entities.3 This principle, known as the sthdnivadbhdva, is of considerable importance for the understanding of nirvacana analysis according to a substitutional model. On the assumption that a nirvacana offers the etymology of a given word, and that there can only be one correct etymology of any one given word, scholars have striven to ascertain which is the correct or preferred alternative when more than one have been suggested. Such an approach is simply a red herring, to which the nirvacanas of 'Bhairava' convincingly testify. To my mind, there can be no doubt that Yaska too states equally valid alternatives when he offers more than one nirvacana for any given word. Most recently this issue has been discussed by Mehendale (1986) when criticising the view expressed by Bronkhorst (1981:6-7): 'On many occasions the Nirukta gives several etymologies of one single word in one single meaning. There is reason to believe that all such etymologies were considered simultaneously correct.' Mehendale is of a different opinion (1986:119): 'His (= Yaska's) use of such expressions as vd, api vd, yad vd while giving alternative etymologies for a given word - and this is also true in the case of the word
3

See p. 188 above.

270

Indian semantic analysis

nighantu - clearly shows that in his opinion these are possible alternatives and it is presumed that a new suggestion cancels the ones made previously.' In my view this presents no tenable argument at all: it is perfectly possible to argue that vd, api vd, and yad vd simply suggest alternatives with no preference for any particular one of them. And, understandably, Mehendale admits to some puzzlement as to why Yaska gives an additional derivation of anna from dAnam when its derivation from A ad 'is clear' (ibid.): It is true that it is not easy to say why Yaska does this. It is possible that he did this under the influence of the Brahmanas and the Upanisads where, occasionally, words of known derivation, have been derived in an 'unorthodox' way, witness, for example, the derivation of yajna in the Sat. Br. 3.9.4.23, Ch. Up. 4.16.1; 8.5.1, or of yajus in the Br. Up. 5.13.2. Or one may say that Yaska did it because to him the derivation of anna from ad- was not that obvious since not all the roots which end in -d yield similar formations, and even in the case of roots like bhid-, nud- etc. alternative forms are available. Be that as it may, a case like this cannot be an argument to infer that Yaska believed in the simultaneous correctness of alternative derivations. There seems to be only one way of understanding the simultaneous correctness of multiple derivations of nighantu. We have to suppose that at one time the samdmndya was called nigantu, samdhantu and samdhartu and that in course of time all these designations assumed one identical form nighantu. Only this way the three derivations become simultaneously correct.4 Mehendale goes on (ibid.): 'Bronkhorst rightly discards this assumption, and yet says that Yaska believed in the simultaneous correctness of the alternative derivations.' In my view Bronkhorst is absolutely right, and the explanation suggested by Mehendale, although he discards it, is indeed not the only way one could explain how numerous derivations of a single word all can be considered valid. I am, however, slightly reluctant to use the* expression 'simultaneously correct' and would rather describe them as equally correct in that one can only invoke one alternative at any given moment. Withki the timeless framework of a substitutionafmodel alternative nirvacanas are equally valid. They are simply alternative placeholders in semantic space. In vdrttika 12 on A 1.1.56 the question is raised whether the sthdnivadbhdva principle is incompatible with the doctrine that speech elements are eternal (nitya): anupapannam sthdnyddesqtvam nityatvdt, 'the condition of being sthdnin and ddesa does not come about because [of the doctrine that speech elements] are eternal'. Pataiijali remarks on this as follows (Mbh 1:137,4-6): sthdnyddesa ity etan nityesu sabdesu, nopapadyate I kim kdranam I nityatvdtl sthdnT hi ndma yo bhiltvd na bhavatil ddeso hi ndma yo 'bhutvd bhavatil etac ca nityesu sabdesu nopapadyate yat sato ndma vindsah sydd asato vd prddurbhava itill
4

The reference is to Nirukta 1.1: nighantavah kasmatl. . .1 te nigantava eva santo nigamand nighantava ucyanta ity aupamanyavah I api vd hanandd eva syuh I samdhatd bhavanti I samdhrtd bhavanti I

Epilogue

271

This [condition of being] sthdnin and ddesa does not come about when speech elements are eternal. Why? Because of [the doctrine of] eternality. For what we call a sthdnin is something which, having been, is no longer, and what we call an ddesa is something which, not having been [previously], comes to be. And this does not come about when speech elements are eternal, namely that something existent should be destroyed, or something [previously] non-existent become manifest. Kaiyata comments (MbhP I:453a): ekadesavikdre ypi yatra nityatvahanis tatra sarvavikdre kuto nityatvam, 'even when you have change in just a part the principle of eternality is violated, so how can you have eternality when there is change of the whole?'. He adds: yatrdnvayo 'pi kasya cin ndstity arthah, 'that is to say, when there is not even the continued presence of anything'. When, for example, the verbal suffix -ti is replaced by -tu there is at least the continued presence of t. This is not the case when there is a change of the whole, so, it is argued, in that case the eternality principle is definitely violated. Nagesa makes this clear (MbhU I:453b): sarvavikara itil aster bhur ityddaull yo bhutva ityddi bhdsyasya prdgabhdvdpratiyogitve sati dhvamsdpratiyogitvam nityatvam iti tdtparyamll 'When there is change of the whole': [that is to say,] in cases such as [A 2.4.52] aster bhuh. The intention of the Bhdsya beginning 'that which having been' is that [for something] to be eternal is [for it] to have an existence which, given that it is not the counterpart of prior non-existence, is not the counterpart of [posterior] destruction [either]. Thus, to be eternal is to have a kind of existence which is qualified neither by prior non-existence nor by posterior destruction. The argument that this is incompatible with the condition of being sthdnin and ddesa is refuted by vdrttika 13: siddham tu yathd laukikavaidikesv abhutapurve 'pi sthdnasabdaprayogdt, 'but [the doctrine of nityatva] is properly managed because the word sthdna is used with reference to [something] which was not [there] previously5 as well, just as in everyday speech and in Vedic'. Patanjali remarks (Mbh 1:137,8-11): siddham etatl kathaml yathd laukikesu vaidikesu ca krtdntesv abhutapurve 'pi sthdnasabdo vartatel loke tdvad upddhydyasya sthdne sisya ity ucyate na ca tatropadhydyo bhutapurvo bhavatil vede 'pi somasya sthdne putikatrndny abhisunuydd ity ucyate na ca tatra somo bhutapurvo bhavatilI This is properly managed. How? The word sthdna occurs also with regard to [something] which was not there previously, just as in mundane and Vedic precepts.6 For example, in everyday speech it is said The pupil instead of the teacher.' But here the teacher is not [someone] who was [there] previously. In the Veda also. It is said 'The Putlka plant should be pressed instead of Soma.' But here Soma is not [something] which was [there] previously.
5 6

Patanjali will later (Mbh 1:137,15) say that the word sthdna occurs with regard to that which has been (Jbhutapurve) as well {bhutapurve cdpi sthdnasabdo vartate); see below. This translation of krtanta follows Joshi and Roodbergen 1986:121, note 492.

272

Indian semantic analysis

Within a substitutional model the teacher and the pupil, the Soma and the Putlka plants are simply alternative placeholders without any notion of earlier or later. The teacher does not cease to exist, although the pupil appears in his place. Similarly, within grammatical space, there is no violation of the doctrine of nityatva. The non-temporal substitution is proved by referring to the meaning of the word sthdna in everyday and ritual contexts where the sthdnin is not former. i A further argument is brought forward by vdrttika 14: kdryaviparindmdd vd siddham, 'or it is properly managed because of a transformation of that which is to be formed'. This understanding of kdryaviparindmdt is based on Patanjali who explains (Mbh 1:137,13-14): kdryd buddhih sd viparinamyate, 'a cognition which is to be formed, that is what undergoes a transformation'. Kaiyata comments (MbhP I:453b): anyasydtra viparindmdsambhavdt sdmarthydt kdryatvena buddhir grhyate na tu kdryasabdo buddhiparydyah, 'because it is impossible for anything else to undergo the transformation here, cognition (buddhi) is understood on the strength that it is what is to be formed (kdrya), but the word kdrya is not a synonym of buddhi". Patanjali explains the purpose of the vdrttika (Mbh 1:137,17-26): parihdrdntaram evedam matvd pathitam katham cedam parihardntatam sydtl yadi bhutapurve sthdnasabdo vartatel bhutapurve cdpi sthdnasabdo vartatel katham I buddhydl tadyathdl kas cit kasmai cid upadisati prdcinam grdmdd dmrd itil tasya sarvatrdmrabuddhih prasaktdl tatah pascdd aha ye kslrino "'varohavantah prthuparnds te nyagrodhd itil sa tatrdmrabuddhyd nyagrodhabuddhim pratipadyatel sa tatah pasyati buddhydmrdms cdpakrsyamdndn nyagrodhdms cddhiyamdndnl nityd eva ca svasmin visaya dmrd nityds ca nyagrodhd buddhis tv asya viparinamyate I evam ihdpy astir asmd avisesenopadistah I tasya sarvatrdstibuddhih prasaktdl so yster bhur bhavatity astibuddhyd bhavatibuddhirripratipadyate I sa tatah pasyati buddhydstim cdpakrsyamdnam bhavatim cddhtyamdnam I nitya eva ca svasmin visaye 'stir nityo bfiavatir buddhis tv asya viparinamyate II This [vdrttika] has been read on the-assumption that it presents a totally different solution to the problem, but how would it be a totally different solution? [It would] if the word sthdna occurs with regard to [something] which was there previously, and the word sthdna does occur with regard to [something] which was there previously as well. How? By [referring to] a cognition. Just as someone points out to someone else: 'there are mango trees east of the village'. That [person] will have the cognition of mango trees everywhere [east of the village]. Then later he (the speaker) says: 'those [trees] with a milky juice, which grow downward, with broad leaves, they are banyan trees'. Then he (the listener) obtains a cognition of banyan trees through the cognition of mango trees. Then he sees that the mango trees are removed and the banyan trees are established in his mind,7 but the mango trees are in fact permanently established in their own domain, and the banyan trees are permanently established [in theirs]. Only his cognition is transformed. Similarly
7

There is no clear distinction between kartr, karana, and adhikarana here: it happens m the mind by force of the mind.

Epilogue

273

also in this case: [the verbal root] as has been taught to him without specification, so he will have a cognition of as everywhere.8 He then obtains the cognition of [the verbal root] bhu through the cognition of as by [A 2.4.52 which teaches that] bhu appears in the place of as [in the domain of drdhadhdtuka suffixes]. Then he sees that as is removed and bhu is established in his mind, but as is in fact permanently established in its own domain, and bhu is permanently established [in its]. Only his cognition is transformed. So there is no violation of the doctrine of nityatva in this case either. Once the speech elements of the Sanskrit language are considered eternal, one element does not cease to exist when a complementary element appears in its place. The idea of succession, the temporal priority of the sthdnin, is denied. There is no earlier or later, except in the mind. Or, in the words of Kaiyata (MbhP I:454a): na ca nityatvavirodhahl buddhidharmdh kevalam arthesu pratipattrvasdd dropyante na tu tadvasdd arthdndm tathd 'vasthdnam, 'and so there is no violation of [the doctrine of] eternality; it is only the properties of cognitions that are imposed upon things by force of the cogniser, but it is not by such power that the condition of things is [established] as such'. This leaves us with a model of substitution which does not involve any notion of time. If the ddesa is considered like the sthdnin in accordance with the sthdnivadbhdva principle, and there is no notion of temporality involved in the substitutional model, it becomes easy to explain how alternative nirvacanas of a given word in a single meaning can be equally valid. They are simply alternative placeholders in semantic space. There is no prior or posterior, there is no one alternative which has to be preferred. Since the relation between word and meaning is eternally established too, the various alternatives are permanently invokable ideal expressions which the actual forms we meet with in language replace. The temporality problem may even hint at a possible solution as to why a substitutional model could have been adopted in the first place. As noted, in the Atharvaveda and in the Satapathabrdhmana the mythical activity given as the ground for using a name is expressed in a past tense, thus referring to a specific event.9 If a Nirukta formula of the -tehl-eh or iti satah type is interpreted according to a substitutional model, no temporality is expressed at all. This could be the result of an exegesis which suppresses time in order to secure the eternality of the Vedas. This idea of eternality conflicts with the Aitihasikas, the legendarians, who refer to real events and persons in the past. The Nairuktas, however, prefer allegorical interpretations, and, in general, dissolve historicity.10 The temporality and personality problem became really serious for the
8 9 10

That is to say, in the domain of drdhadhdtuka suffixes as well as that of sdrvadhdtuka suffixes. See p. 27 above. Cf. p. 27 above; an example was adduced from Nirukta 2.16 concerning the identity of Vrtra: tat ko vrtrahl megha iti nairuktdhl tvdstro 'sura ity aitihdsikdh, 'but who is Vrtra7 a cloud according to the Nairuktas; an Asura, son of Tvastr, according to the Aitihasikas'.

274

Indian semantic analysis

Mlmamsakas* the extreme Vedic ritualists. For them even myth was arthavdda, a non-injunctive 'explanation of the meaning', an elaboration of the topicmatter coming up in ritual, designed to encourage or dissuade. The purpose of myth was no longer to narrate events of the past but to promote a rite. Already Jaimini argued that the Veda is not composed by anyone and so is eternal.11 If the Veda did not exist before the persons referred to in myths, then it would not be eternal. In fact, any narrative at all would prove that the Veda is not eternal. With an allegorical interpretation and within a substitutional model the personality and temporality problems disappear. One further issue arises as a result of the sthdnivadbhdva principle. If the ddesa is to be considered like the sthdnin, then how do we consider a case where no sthdnin is mentioned? In the context of vydkarana, Vaidyanatha pointed out that since linguistic units occur in complementary distribution, the rules of grammar serve to establish a fixed distribution in this respect. This is possible, Vaidyanatha says, because sthdnins and ddesas both exist in the world, a fact which implies that one has to consider usage.12 In nirvacanasdstra we have to infer the sthdnin which, although not the form met with in actual linguistic usage, has its characteristics transferred onto the ddesa, the form we meet with in actual usage. Still, the sthdnin too is part of the Sanskrit language. It is just a more composite linguistic unit, in its fullest form simply a sentence. In vydkarana substitutes are considered substitutes of "whole words (padas)13 as opposed to the earlier framework of the Pratisakhyas, for example, TaittirTya Prdtisdkhya 1.56: varnasya vikdralopau, 'change and elision are of a speech sound'.14 W. Whitney (1871:40) remarks on the latter: *That is to say, not of a whole word. Where, as by v.19, more than one letter is omitted, each is specified.' In the terminology of vydkarana a pada or word is defined by A 1.4.14 suptinantam padam, 'that which ends in a nominal ending or a verbal ending is &pada\ Panini does not define the notion of a sentenced or vdkya, but various definitions have been suggested by subsequent Paninlyas. The briefest one is offered by Katyayana at A 2.1.1, vdrttika 10: ekatin, 'that which contains a finite verb'.15 Other traditions, most notably Buddhism and the Tantric traditions concerned with the use of mantras consider a pada a larger unit not delimited by word boundaries. The Mantrasastric use of the term pada has been well put by Abhinavagupta in the Tantrdloka discussing the embodiment of the divine consciousness in > 11 12 13 Cf Mlmamsasutra 11 27-32 See p 233 above See p 187 above. 14 See p 175 above 15 This involves the assumption that a finite verb such as asti underlies nominal sentences, a view accepted by Patanjah and the later tradition Patafijall's own view is, roughly, that there must be a verb in a sentence, but that there can be more than one verb Most recently the question of Paniman syntax and the changing notion of sentence has been discussed by Deshpande 1987, 1991 As Deshpande points out (1987.71, 1991 34), Panini held the view that there can be purely nominal sentences and that their derivation does not require the assumption of an underlying finite verb

Epilogue

275

the pan-Saiva Vyomavyapimantra of eighty-one padas {ekdsitipada devi)16 (TA 6.228c-30b): tathd tathd pardmarsasakticakresvarah prabhuh II sthulaikdsitipadajapardmarsair vibhdvyate I tata eva pardmarso ydvaty ekah samdpyate II tdvat tatpadam uktam na suptinniyamayantntam I The omnipotent governor of the totality of Powers [constituted by] the sum of all ideas of whatever form is realised through [the contemplation of] the ideas which arise [one by one] m these eighty-one padas, finite though they are Therefore we define the pada as that unit [of speech] which accommodates the complete idea, not [as m vydkarana] as that which is bounded by a sup or a tin. Thus, as Jayaratha points out by way of illustration on these verses (TAV 4[6] 187,13—16), the six syllables om narnah sivdya are not three padas but one. The use of the term pa da in this sense does not stem from Abhinavagupta but goes back to the Saivagamas.17 A similar use of the term pada is met with in Buddhist texts. Vasubandhu states in his Abhidharmakosabhdsya (AKBh 270,5-6) :18 vdkyam padam, ydvatd 'rthapansamdptih, tad yathd — anityd bata samskdrdh ityevamddi, yena knydgunakdlasambandhavisesd gamyante I [The term] vdkya [of AK 2.47] means pada, insofar as it allows for the complete expression of a meaning, for example, 'impermanent are the samskdras\ etc , [and pada is defined as that] by which the specific relations of action, quality, and time are understood. A syntactic unit is a pada, not necessarily a word which for the grammarian was the natural unit when describing language. In nirvacanasdstra these units may vary. A unit may be a word or contain several words, but the important thing is that the substitution affects the totality. It seems the situation may be viewed as follows. Every noun analysed by a nirvacana is linked to an action, referred to by means of a verbal root or a finite verb. Ultimately this entails that a noun is to be interpreted with regard to kdraka, which involves a finite verb. If it is interpreted in terms of bhdva, one may argue that a finite verb such as asti underlies the analysis. This implies that a noun may be the replacement met with in the real language of a semantically eqiiivalent expression which formally is a sentence. It follows that in nirvacana analysis a noun ultimately embeds a sentence. The stdnivadbhdva principle secures that the characteristics of the sentence are transferred to the noun. It is possible to pursue this line of thinking a bit further. A 1.1.62 pratyayalope pratyayalaksanam teaches that when a suffix is elided and replaced by
16 17

For this mantra see Brunner-Lachaux 1977 240, note 192 By way of example, Matangaparamesvardgama 131 ab (ed Bhatt 1977 241) ekdsitipada dev yd sd saknh sivdtmikd, describing the same Vyomavyapimantra 18 See Simonsson 1982 540-1

276

Indian semantic analysis

zero, the mark of the elided suffix remains to take effect in a subsequent operation caused by the suffix.19 Considering the nature of natural ellipsis in Panini, Deshpande (1985a:36) suggests that in view of A 1.1.62 and the situation of word-ellipsis and its /connection with semantics and syntax, one may formulate a theory that Panini does have a general rule padalope padalaksanam. In the words of Deshpinde (ibid.): 'When a word is zeroed (by natural ellipsis), the zero may still be treated like the original word it replaces.' The existence of such a rule woujd allow us to explain all sorts of elliptical sentences in Sanskrit in accordance with Panini's system. Deshpande concludes (ibid.): However, this principle is restricted to semantics and syntax, and does not extend into the realm of phonological and accentual rules.20 One can justify such a conclusion based on Panini's rules. Panini clearly says that the lost word is the sthdnin 'substituendum' for the deletion or zero (lopa), cf. P. 1.4.105 and P. 2.3.14. Thus, the general rule P. 1.1.56 (sthdnivad adeso ynal-vidhau) should apply to a word and its replacement by a zero. A 1.4.51 akathitam ca teaches that the name karman is assigned also to that kdraka which is unspecified, that is, which has not otherwise been assigned to any specific kdraka category. According to the tradition it is this rule which accounts for ditransitive constructions of the type mdnavakam panthdnam prcchati, 'he asks the boy the way'.21 Kiparsky (1982:39 ff.) has suggested that A 1.4.51 akathitam ca in fact accounts for elliptical constructions where the object is omitted from the sentence and therefore identified as akathita 'not expressed'. He concludes (ibid.:44): 'It seems, then, that Panini assumed a general deletion rule, operating freely in the grammatical derivations of sentences, which drops words which are obvious from the meaning or context.' Let these observations serve to indicate that it is perfectly possible to view the mechanics of nirvacana analysis within a similar framework, yhe word that is explained embeds a sentence with most of its elements elided and replaced by zero. It is still possible to treat the word like the original sentence it ultimately replaces. Indeed, this may suggest a way of thinking not confined to vydkarana alone. It may even provide a clue to the understanding of the practice of assigning metaphysical meaning to all the sounds of the Sanskrit language resorted to by certain Tantrlc traditions. To conclude, I have argued that an analysis of the Hi satah type conveys the following information: there is (sat-) an x suehmat it is true (sat-) to say of it
19 20

21

This rule and Patanjah's remarks on it have been discussed by Benson 1990,141-68. As far as I can see, this is not entirely correct. There are some accent rules where lopa is used to indicate that a whole word is elided. A 8.1.62 cdhalopa evety avadhdrane teaches that when there is elision of the words ca or aha and the word eva is used in their place to denote a restriction, the first of the finite verbs retains its accent. A 8.1.63 cddilope vibhdsdlea,ches that when there is elision of the words ca, vd, ha, aha, or eva the first of the finite verbs optionally retains its accent See Benson 1990 134 for a discussion of these rules in the cdntext of A 1.1 60 adarsanam lopah S.D Joshi (1976) has argued that it is the preceding rule A 1 4.50 tathdyuktam cdnlpsitam and not A 1.4 51 which accounts for double accusative constructions of this type.

Epilogue

277

'it is x-ing' or 'it is being x-ed' etc., and therefore it is called 'x\ The fact that the sthdna, the location in grammatical space, is identified as artha opens new perspectives for the interpretation of an Hi satah type of analysis in the light of a substitutional model. Since the term artha frequently refers to that which is designated by a name as well as to the meaning of that name, it is less crucial whether we take artha in the sense of linguistic meaning or we take it in the sense of the thing meant. Once the genitive of satah is interpreted as a sthdnasasthi and sthdna is identified as artha, then an analysis such as megho mehatiti satah can be interpreted: 'meghah' occurs in the meaning of that which really exists such that it is true to say of it: it rains. Such an interpretation provides no difficulties and is supported by an investigation of the substitutional model in the literatures of grammar and ritual. Basically, then, a nirvacana analysis works so that it establishes the truth-value of a sentence in the sense that if a true sentence can be established as an ideal version of a prevalent term, then that term is infused with the meaning of the true sentence. As such, nirvacana analysis is an empirical theory of what expressions mean by force of its appearance as a theory of truth. The meaning of a sentence is intimately connected with whether it is true or false. The sentence 'Snow is white' is true if and only if snow is white. While discussing the nirvacanas of 'Bhairava' I pointed out that one can give exactly the same truth conditions in a language by making systematic shifts in what the singular terms refer to and what satisfies the predicates.22 One gets a different thought if one gets a different sentence to represent it. And whether we can say of any particular that it has necessary or contingent properties depends simply on the way it is described. The important feature of nirvacana analysis is that the attribution of properties happens through the name itself, for example 'Bhairava'. The analysis establishes an ideal expression which has the semantic power to identify the nature of that which a name signifies. This, I suggest, is how it came to be a powerful tool in cultural discourse. If someone has the power to determine the reference of the words of a shared language, then that person has the capacity to interpret and determine the contents of thoughts. If knowledge is justified true belief, then it would seem that all the true beliefs of a consistent believer constitute knowledge. A nirvacana analysis justifies a certain belief, and this justification implies that the belief is true (sat-). A nirvacana analysis is thus an epistemological device. The notion of reference in modern semantics is a theoretical concept we introduce in order to have a satisfactory theory of what sentences do, and a theory of truth is a theory about semantic relations between sentences in a language. Obviously, I am not claiming that nirvacanas work 6r are to be interpreted within such a framework. There is a big difference between the question of what makes sentences true in a philosophically acceptable way, and the way sentences are made true by consent in the community. This alone makes it clear
22

See pp. 3 and 97 above

278

Indian semantic analysis

that the nirvacana device is very different from modern semantic theories of truth or truth-based theories of semantics in that it lacks explanatory power in the philosophical sense. In the light of the technical framework of the Nirukta it is of course also possible to adopt a different model for the interpretation of nirvacanas. On the view that the -teh and -eh forms are ablatives one would simply face statements that single out the verbal element which underlies the grammatical formation and identify the action or event considered the reason for a particular name. This is a possibility also if they are considered genitive forms, and it is no longer crucial to determine which case ending we are dealing with. But this does not exclude an interpretation according to a substitutional model either. Even within the AstadhyayT one and the same genitive form is to be interpreted in different ways in different contexts.23 As far as the material from the Nirukta is concerned, we are dealing with elliptical ways of expression*which lend themselves to various interpretations according to the different levels and historical presuppositions of Yaska's technical language. There is no reason why these interpretations cannot be simultaneously operative. Indeed, the situation reflects rather strikingly the fundamental debate on the nature of words or speech elements. One view holds that they are eternally established (nitya), while the alternative view holds that they are produced (kdrya), as if processed by grammar.24 It seems worth considering once again the Brahmana equations mentioned earlier, equations of the type samvatsaro yajftah, 'the sacrifice is the year' (SatBr 11.2.7.1), or ddityo 'gnih,{ 'the [sacrificial] fire is the sun' (SatBr 10.5.4.1).25\Here 'the sacrifice' or 'the fire' would be termed ddesa. They are ritual elements which replace cosmological entities. It is in these identification formulas that we can see a move towards technical nirvacanas. In the context of ritual sthdna could be the levels of interpretation known as'adhidevatd, adhydtman and adhiyajna, levels which establish a relation between macrocosm, microcosm and ritual. This way of interpreting is very common in the l Brahmanas and Upanisads. With the sthdna identified as artha 'meaning' a nirvacana is to be interpreted not only as 'X in the place of Y' but as 'X in the meaning of Y \ This implies that we have an ideal expression Y which is the placeholder in semantic space. This place, that is to say, this meaning, is instead occupied by the item X which replaces Y. Knowledge of tharsecret place or meaning and that secret entity Y is provided by experts who are in command of the^ools of the tradition. In this way change is achieved through substitution in that new meaning may be encoded into old terms by means of a substitutional model. I have stated already that this is not the only way to look at nirvacana analysis,
23

24

By way of example, the genitive of A 6,4.1 angasya is to be interpreted as a sthdnasasthT, a substitutional genitive, in A 7 2 102 tyadpdinam ah, but as an avayavasasthT, a partitive genitive, when it recurs by25 anuvrtti in A 7 2 106 tadoh sah sdv anantyayoh, see p 226 above See p 186 above. See p 180 above. ' '

Epilogue

279

but it has the advantage of showing the flexibility of the various types of analysis while providing a superstructure that encompasses them all within a model known from other areas of Sastric Sanskrit, notably the literatures of grammar and ritual. The model arises out of the wealth of material provided by the Indian tradition itself, and it has the advantage of directly involving the artha of a given term, the very goal nirvacanasdstra strove to achieve.

ABBREVIATIONS AND BIBLIOGRAPHY

Primary sources which are arranged in volumes, chapters, or verses are referred to by means of these. Otherwise references are to page and line. A Astddhydyi of Panini. Ref. to ed. O. Bohtlingk 1887. A ASS Ananddsrama Sanskrit Series, Poona. Abhyankar, Kashinath Vasudev (tr.): 1960, The Paribhdsendusekhara ofNdgojfbhatta, edited and explained by F. Kielhorn; part 2: translation and notesy Poona: BORI. (New edition, with preface by K.V. Abhyankar, of Kielhorn 1871-4.) (ed.): 1962 The Paribhdsendusekhara ofNdgojibhatta, edited critically with the commentary Tattvddarsa of MM. Vasudev Shastri Abhyankar; part 1 Srimadupddhydyopandmakasivabhattasutasatlgarbhajandgojibhattakrtah srimanmahdmahopddhydydbhyamkaropdhvavdsudevasdstrTviracitaydtattvddarsdbhidhayd vydkhyayd samavetah paribhdsendusekharah, prathamo bhdgah), Poona: BORI. (New edition, with new Sanskrit commentary, of Kielhorn 1868.) (ed.): 1967, Paribhdsasamgraha {a collection of original works on vydkarana paribhdsds) (Paribhdsdsamgrahah (Vyddisdkatdyanddiproktaparibhdsdpdthdndm tatpranitavrttindm ca samgrahah)) {BORI Post-graduate and Research Department Series 7), Poona: BORI. Abhyankar, Kashinath Vasudev and Vishnu Prabhakar Limaye (eds.): 1965, VdkyapadTya ofBhartrhari (University of Poona Sanskrit and Prakrit Series 2), University of Poona. ABS Ahirbudhnyasamhitd, ed. by M.D. Ramanujacarya under the supervision of F. Otto Schrader, 2 vols., Madras: Adyar Library, 1916. (Revised edn by V Krishnamacharya (Adyar Library Series 4), Madras: Adyar Library, 2nd edn 1966.) ' Agrawala, VS.: 1962, 'GawT, in Indological studies in honor of W.N. Brown (American Oriental Series 47), ed. by E. Bender, New Haven: American Oriental Society, pp. 1-7. , . AGS Asvaldyana Grhyasutra, ed. with the commentary of Gargya Narayana by Ramanarayana Vidyaratna and Anandacandra Vedantavaglsa (Bibliotheca Indica 57), Calcutta: Asiatic Society of Bengal, 1866-9. AK(Bh) Abhidharmakosa(-bhdsya) of Vasubandhu, ed. with the Sphutdrthd commentary of Yasomitra by Swami Dwarikadas Shastri, vol. 1 (Bauddha Bharati Series 5), Varanasi: Bauddha Bharati, 1970. Aklujkar, Ashok: 1989, 'Sambandha and abhisambandha', Journal of Indian Philosophy 17: 299-307. ApSS Apastamba Srautasutra, ed. with the commentary of Rudradatta by Richard

Abbreviations and bibliography

281

Garbe, 3 vols. (Bibhotheca Indica 92), Calcutta: Asiatic Society of Bengal, 1882-1902. Apte Revised and enlarged edition of V.S. Apte's The practical Sanskrit-English dictionary, eds.-in-chief RK. Gode and C.G. Karve, 3 vols., Poona: Prasad Prakashan, 1957-9. ASS Asvaldyana Srautasutra, ed. with the commentary of Gargya Narayana by Ramanarayana Vidyaratna (Bibhotheca Indica 49), Calcutta: Asiatic Society of Bengal, 1864-74. (Repr. 1989.) Aufrecht, Theodor (ed.): 1859, Ujjvaladatta's commentary on the Unddisutras, Bonn: Marcus. AV Atharvavedasamhitd, ed. by R. Roth and W.D. Whitney, 2nd edn, revised by M. Lindenau, Berlin: Ferd. Dummler, 1924. BAUp Brhaddranyakopanisad, ed. by KasTnatha Sastrl Agase (AASS 15), Poona: Anandasrama, 4th edn 1939. BD Brhaddevatd of Saunaka, ed. by Arthur Anthony Macdonell (HOS 5), Cambridge, Mass.: Harvard University Press, 1904. Benson, James William: 1990, PatanjaWs remarks on ariga (Oxford University South Asian Studies Series), Delhi: Oxford University Press. Bhagavaddatta: 1931, Vaidika vdnmaya kd itihdsa, dusrd bhdga, vedom ke bhdsyakdra (vol. 2: the commentators of theVedas), Lahore. (English trans, by Satyashrava: A comprehensive history ofVedic literature, New Delhi: Pranava Prakashan, 1977.) Bhaglratha Prasada TripathI: 1965, PdninTya-dhdtupdtha-samTksd (Sarasvati Bhavana Studies (SarasvatTbhavana-adhyayanamdld) 14), Varanasi: Varanaseya Sanskrit Vishvavidyalaya (Varanaseyasamskrtavisvavidyalayah). (In Sanskrit.) Bhate, Saroja: 1968, 'Some primary and secondary suffixes known to Yaska', Journal of the University of Poona 27: 121-32. (Repr. PCASS, Class A, 15.) Bhatt, N.R. (ed.): 1977, Matangaparamesvardgama (Vidydpdda) avec le commentaire de Bhatta Rdmakantha (PIFI56), Pondichery: PIFI. Bhattacharya, Bishnupada: 1950, 'Niruktavdrttika - a lost treatise', Indian Historical Quarterly 26: 159-65. (Repr. Bhattacharya 1958: 96-109.) 1958, Ydska's Nirukta and the science of etymology, Calcutta: K.L. Mukhopadhyay. BHSD Franklin Edgerton, Buddhist Hybrid Sanskrit Dictionary, New Haven: Yale University Press, 1953. (Repr. Delhi: Motilal Banarsidass, 1970.) BhSS Bhdradvdja Srautasutra, ed. with the Paitrmedhika and Parisesa Sutras by C.G. Kashikar, Poona: Vaidika Samsodhana Mandala, 1964. Biardeau, Madeleine and Charles Malamoud: 1976, Le sacrifice dans Vlnde ancienne (Bibliotheque de VEcole des Hautes Etudes, Section des Sciences Religieuses 79), Paris: Presses Universitaires de France. Bohtlingk, Otto (ed.): 1887, PdninVs Grammatik, herausgegeben, ubersetzt, erldutert und mit verschiedenen Indices versehen, Leipzig: Haessel. (Repr. Hildesheim: 01ms, 1964,1971.) BORI Bhandarkar Oriental Research Institute, Pune. BPD Bodhapancadasikd by Abhinavagupta with commentary (-vivarana) by Pt. Harabhatta (KSTS 76), ed. by Pt. Jagaddhara Zadoo, Srinagar, 1947. Bronkhorst, Johannes: 1981, 'Nirukta and Astadhyayl: their shared presuppositions', 7/7 23:1-14. 1984, 'Nirukta, Unadi Sutra, and Astadhyayl: a review article', //J27: 1-15. (Review article of Mehendale 1978.)

282

Abbreviations and bibliography

1985, 'A possible quotation from the Niruktavarttika known to Durga in the Yuktidlpika', in Proceedings of the Fifth World Sanskrit Conference, ed. by R.N. Dandekar and RD. Navathe, New Delhi: Rashtriya Sanskrit Sansthan, pp. 90-100. 1986, Tradition and argument in classical Indian linguistics: the bahirangaparibhdsd in the Paribhdsendusekhara (Studies of Classical India 6), Dordrecht: Reidel. 1987, Three problems pertaining to the Mahdbhdsya (BORI Post-graduate and Research Department Series 30, 'Pandit Shripad Shastri Deodhar Memorial Lectures' (third series)), Poona: BORI. 1990, lvdrttika\ Wiener Zeitschrift fur die Kunde Sudasiens 34: 123-46. Brunner-Lachaux, Helene (ed., trans., annot.): 1963, Somasambhupaddhati, premiere partie: le rituel quotidien dans la tradition sival'te de VInde du Sud selon Somasambhu (PIFI 25.1), Pondichery: Institut Francais d'Indologie. 1977, Somasambhupaddhati, troisieme partie: rituels occasionnels dans la tradition sivalte de VInde du Sud selon Somasambhu, II: diksd, abhiseka, vratoddhdra, antyesti, srdddha (PIFI 25.Ill), Pondichery: Institut Frangais d'Indologie. BSPS Bombay Sanskrit and Prakrit Series. BSS Bombay Sanskrit Series. BSS Baudhdyana Srautasutra, ed. Willem Caland, 3 vols. (Bibliotheca Indica 163), Calcutta: Asiatic Society of Bengal, 1904-23. Buddruss, Georg (ed.): 1971, Paul Thieme; Kleine Schriften (Glasenapp-Stiftung 5,1-2), 2 vols. (with continuous pagination), Wiesbaden: Steiner. (2., unveranderte Auflage mit einem Nachtrag 1984 zur Bibliographic, 1984.) Caland, Willem: 1903: Uberdas rituelle Sutra des Baudhdyana (Abhandlungenfur die Kunde des Morgenlandes 12,1), Leipzig: Brockhaus. Cardona, George: 1967, 'Panini's syntactic categories', Journal of the Oriental Institute, Baroda 16: 201-15. 1967-8, 'Anvaya and vyatireka in Indian grammar', Adyar Library Bulletin 31-2 (V. Raghavan Felicitation Volume): 313-52. 1968, 'Panini's definition, description, and use of svarita\ in Pratiddnam: Indian, Iranian and Indo-European studies presented to Franciscus Bernardus Jacobus Kuiper on his sixtieth birthday, ed. by J.C. Heesterman, G.H. Schokker, and V.I. Subramoniam, The Hague-Paris: Mouton, pp. 448-61. * 1969, Studies in the Indian grammarians I: the method reflected in the sivasutras (Transactions of the American Philosophical Society,* new series 59. 1), Philadelphia: American Philosophical Society. 1970, Review of Staal 1967, /// 12: 232-9. 1970a, 'Some principles of Panini's grammar', Journal of Indian Philosophy 1: 40-74. ' \ 1974, Tanini's karakas: agency, animation and identity', Journal of Indian Philosophy 2: 231-306. 1974a, 'On Panini's metalinguistic use of cases', in Charudeva Shastri felicitation volume, ed. committee S.K. Chatterji, V. Raghavan, R.N. Dandekar, Vishva Bandhu, A.D. Pusalkar; K.A.S. Iyer, and Satya Vrat Shastri, Delhi: Charudeva Shastri Felicitation Committee, pp. 305-26. 1976, Pdnini: a survey of research (Trends in Linguistics, State-of-the-Art Reports 6), The Hague-Paris: Mouton. (Repr. Delhi: Motilal Banarsidass, 1980.) 1977, Review of PatahjalVs Vydkarana-Mahdbhdsya Bahuvnhidvandvdhnika

Abbreviations and bibliography

283

(P. 2.2.23-2.2.38), text, translation and notes by J.A.R Roodbergen, ed. by S.D. Joshi, Poona: University of Poona, 1974, Journal of the Oriental Institute, Baroda 26, 3: 328-42. 1978, 'Still again on the history of the Mahabhasya', Annals of the Bhandarkar Oriental Research Institute 58-9 (Diamond Jubilee Volume): 79-99. 1980, 'On the paribhasa upapadavibhakteh kdrakavibhaktir ballyasV, Studien zur Indologie und Iranistik 5/6: 27-48. 1981, 'On reasoning from anvaya and vyatireka in early Advaita', in Studies in Indian philosophy (A memorial volume in honour of Pandit Sukhlalji Sanghvi), ed. by Dalsukh Malvania and Nagin J. Shah, Ahmedabad: L. D. Institute of Indology, pp.79-104. 1983, Linguistic analysis and some Indian traditions {Post-graduate and Research Department Series 20, 'Pandit Shripad Shastri Deodhar Memorial Lectures' (first series)), Poona: BORI. 1988, Pdnini: his work and its traditions. Volume one: background and introduction, Delhi: Motilal Banarsidass. (Second edn, revised and enlarged, 1997.) Carrithers, Michael, Steven Collins and Steven Lukes (eds.): 1985, The category of the person, Cambridge University Press. ChUp Chdndogyopanisad, ed. by Kaslnatha Sastrl Agase (AASS 14), Poona: Anandasrama, 1913. CNRS Centre National de la Recherche Scientifique, France. D Durga's commentary (Rjvartha) on the Nirukta: The Nirukta of Ydska (with Nighantu) edited with Durga's commentary, vol. I (BSS 73) ed. by H.M. Bhadkamkar assisted by R.G. Bhadkamkar, Bombay 1918, vol. II (BSS 85) ed. by R.G. Bhadkamkar, Bombay 1942. (Ref. to vol., page and line.) Davidson, Donald: 1974, 'On the very idea of a conceptual scheme', Proceedings and Addresses of the American Philosophical Association 67 (1973-4): 5-20. (Repr. Davidson 1984: 183-98.) 1975, Thought and talk', in Mind and language (Wolfson College Lectures 1974), ed. by Samuel Guttenplan, Oxford: Clarendon Press, pp. 7-23. (Repr. Davidson 1984: 155-70.) 1979, 'The inscrutability of reference', The Southwestern Journal of Philosophy 10: 7-19. (Repr. Davidson 1984: 227-41.) 1984, Inquiries into truth and interpretation, Oxford: Clarendon Press. 1986, 'A nice derangement of epitaphs', in Ernest LePore (ed.): Truth and interpretation: perspectives on the philosophy of Donald Davidson, Oxford: Blackwell, pp. 433-46. Deeg, Max: 1995, Die altindische Etymologie nach dem Verstdndnis Ydskas und seine Vorgdnger: eine Untersuchung u'ber ihre Praktiken, ihre literarische Verbreitung und ihr Verhdltnis zur dichterischen Gestaltung und Sprachmagie, Dettelbach: Joseph H. Roll. Des Desopadesa of Ksemendra (KSTS 40), ed. by Madhusudana Kaul Sastrl, Poona, 1923. Deshpande, Madhav M.: 1985, 'Historical change and the theology of eternal Sanskrit', KZ 99: 122-49. 1985a, Ellipsis and syntactic overlapping: current issues in Pdninian syntactic theory (Post-graduate and Research Department Series 24, 'Pandit Shripad Shastri Deodhar Memorial Lectures' (second series)), Poona: BORI.

284

Abbreviations and bibliography

1987, Taninian syntax and the changing notion of sentence', Annals of the Bhandarkar Oriental Research Institute 68: 55-98. 1991, Taninian syntax and the changing notion of sentence', in Hans Henrich Hock (ed.): Studies in Sanskrit syntax: a volume in honor of the centennial ofSpeijer's Sanskrit Syntax (1886-1986), Delhi: Motilal Banarsidass, pp. 30-43. (Earlier and briefer version of Deshpande 1987.) Dhp the Paniniya dhatupdtha. Ref. to ed. O. Bohtlingk 1887. DIksa, Saroja: 1989, Aitareya evam taittinya brdhmanom ke nirvacana, Delhi: Nag Publishers. DPU Unddisutras in the dasapddi recension. Ref. to ed. Yudhisthira Mlmamsaka 1943. . Dwarikadas Shastri (Dvarikadasah Sastrl) (ed.): 1964, The MddhavTyd Dhdtuvrtti (a treatise on Sanskrit roots based on the Dhatupdtha of Pdnini by Sdyandcdrya) (Prachya Bharati Series 1), Varanasi: Prachya Bharati Prakashan. 1971, Bhdsdvrtti (a commentary on Pdnini's grammar) by Purusottamadeva (Snmatpurusottamadevaviracitd bhdsdvrttih (PdniniydstddhydyTvydkhyd)) (Prachya Bharati Series 9), Varanasi: Tara Publications. Evans, Gareth: 1982, The varieties of reference, ed. by John McDowell, Oxford: Clarendon Press. EVP Louis Renou: Etudes Vediques et Pdnineennes, 17 vols. (Publications de VInstitutde Civilisation Indienne, serie in-8° 1,2,4,6,9,10,12,14,16,17,18,20, 22, 23, 26, 27, 30), Paris: Boccard, 1955-69. Filliozat, Pierre^Sylvain (tr.): 1976, Le Mahdbhdsya de Patahjali avec le Pradipa de KcziyatcTet VUddyota de Ndgesa: adhydya 1, pdda 7, dhnika 5^7 (PIFI 54,2), Pondichery: Institut Francais d'Indologie. Frege, Gottlob: 1893, Grundgesetze der Arithmetik, vol. 1, 2nd edh, Jena. (Repr. Hildesheim: Georg Olm, 1962.) Geldner Karl Friedrich Geldner: Der Rig-Veda aus dem Sanskrit ins Deutsche iibersetzt und mit einem laufenden Kommentar versehen, 4 vols. (HOS 33-6), Cambridge, Mass.: Harvard University Press, 1951-7. Gellner, Ernest: 1993, 'What do we need now? Social anthropology and its new global context', in The Times Literary Supplement, 16 July 1993 (No 4711), pp. 3-4. Gombrich, Richard: 1992, 'The Buddha's Book of Genesis?', IIJ 35: 159-78. 1992a, 'Why is a khattiya called a khattiyal The Agganna Sutta revisited', Journal of the Pali Text Society 17: 213-14. Gonda, Jan: 1955, 'The etymologies in the ancient Indian Brdhmanas\ Lingua 5:61-86. (Repr. in J. Gonda: Selected Studies 2, Leiden: Brill, 1975, pp. 32-57.) 1975, Vedic literature (HIL 1,1), Wiesbaden: Harrassowitz. 1977, The ritual Sutras (HJlL 1,2), Wiesbaden: Harrassowitz. Grassmann Hermann Grassmann: Worterbuch zum Rig-Veda, Leipzig: Brockhaus, 1873. Gren-Eklund, Gunilla: 1984, 'Causality and the method of connecting concepts in the Upanisads', Indologica Taurinensia Yl\ 107-18. 1986, 'The cots are crying', in Kalydnamitrdrdganam: essays inxhonour of Nils Simonsson, ed. by E. Kahrs, Oslo: Norwegian University Press, pp. 79-97. Grierson, George Abraham: 1932, A dictionary of the Kashmiri language, Calcutta: Asiatic Society of Bengal. Harris, Roy: 1981, The language myth, London: Duckworth.

Abbreviations and bibliography

285

Heesterman, Jan: 1979, review of M. Biardeau and C. Malamoud 1976, IIJ 21: 47-9. Hel Helaraja's commentary on VP, kdnda 3, part 1, ed. K.A. Subramania Iyer (Deccan College Monograph Series 21), Poona: Deccan College, 1963. HIL A history of Indian literature, ed. by Jan Gonda, Wiesbaden: Harrassowitz, 1975-. Hirakawa, Akira et al.: 1973, Index to the Abhidharmakosabhdsya; part one: SanskritTibetan-Chinese, Tokyo: Daizo Shuppan Kabushikikaisha. HOS Harvard Oriental Series, Cambridge, Mass. IIJ Indo-Iranian Journal. Ikari, Yasuke: 1969, 'A study on the Upanisadic term ddesa\ Indogaku Bukkyogaku Kenkyu (Journal of Indian and Buddhist Studies) 17,2: 684-9. (In Japanese.) JA Journal Asiatique, Paris. JAOS Journal of the American Oriental Society. JosI, Bhargavasastrl Bhikajl (ed.): 1951, Snmadbhagavatpatanjalimaharsipranite vydkaranamahdbhdsye navdhnikam, AstddhydyTprathamddhydyaprathamapddavydkhydnam, srTmadupddhydyakaiyatanirmitapradipaprakdsitam, sarvatantrasvatantrasnmanndgesabhattaviracitoddyotodbhdsitam; (asya vaidyandthakrtacchdydsitaldni catvdry dvartandni m.m. sivadattaraghunathapanditdbhydm sampdditdni) tad etad iddnim kdsivisvavidydlayddhTtasdstrena mumbdpunyavilsanmahdvidydlayddhydpakena parasurdmaksetrdbhijanena josikulabhusanasnbhikdjTsarmaputrena bhdrgavasdstrind sdstrdcdryena vydkarandnubandhigranthddisdhdyyam samavalambyatippanapdthabhedddipradarsanapuraskarena pariskrtam\ pahcamam samskaranam (Patanjali's Vydkaranamahdbhdsya with Kaiyata's Pradipa and Nagesa's Uddyota; vol. 1: adhydya l,pdda 1, dhnika 1-9, edited with footnotes including extracts from Vaidyanatha's Chdyd compiled by pandits Sivadatta (Kudala) and Raghunatha (Sarma), etc., 5th printing), Bombay: Nirnaya Sagar Press. Joshi, Shivaram Dattatray: 1969, Patanjali's Vydkarana-mahdbhdsya avyayibhdvatatpurusdhnika (P. 2.1.2—2.1.49); edited with translation and explanatory notes by S.D. Joshi in collaboration with J.A.F. Roodbergen (PCASS, Class C, 5), Poona: University of Poona. 1976, Tanini's rules 1.4.49, 1.4.50 and 1.4.51', in Centre of Advanced Study in Sanskrit Studies 3 (PCASS, Class E, 4), Poona: University of Poona, pp. 59-70. Joshi, Shivaram Dattatray and J.A.F. Roodbergen (trs.): 1975, Patanjali's Vydkaranamahdbhdsya: kdrakdhnika (P. 1.4.23-1.4.55); introduction, translation and notes (PCASS, Class C, 10), Poona: University of Poona. 1980, Patanjali's Vydkarana-mahdbhdsya: vibhaktydhnika (P. 2.3.18-2.3.45); introduction, text, translation and notes (PCASS, Class C, 12), Pune: University of Poona. 1985, 'On P. 1.1.56', JAOS 105: 469-77. 1986, Patanjali's Vydkarana-mahdbhdsya: paspasdhnika; introduction, text, translation and notes (PCASS, Class C, 15), Pune: University of Poona. 1990, Patanjali's Vydkarana-mahdbhdsya sthdnivadbhdvdhnika: part I (P. 1.1.56-1.1.57): introduction, text, translation and notes (BORI Research Unit Publications 11), Poona: BORI. Joshi, Venkatesh Laxman (ed.): 1966, Praudha Manoramd with commentary Sabdaratna ofHari DTksita, vol. 1 (Deccan College Monograph Series 31), Poona: Deccan College Postgraduate and Research Institute. JV Jainendramahdvrtti of Abhayanandin (Jainendravydkaranam tasya tikd

286

Abbreviations and bibliography

dcdrya-abhayanandipranita jainehdramahdvrttih (JndnapTtha MiirtidevT Jainagranthamdld, samskrta granthdnka 17)), ed. Shambu Nath Tripathi and Mahadeo Chaturvedi, Varanasi: Bharatiya Jnanapltha KasI, 1956. (Ref. to vrtt{ under sutra number.) Kahrs, Eivind G.: 1980, Bidrag til interpretasjonen av Nirukta (Studier i Ydskas tekniske metoder for analyse av ord, med spesiell vektpa analyser av type -anat og iti satah), University of Oslo. (Diss.) (In Norwegian.) 1983, 'Yaska's use of kasmdf, IIJ 25: 231-7. 1984, 'Yaska's Nirukta: the quest for a new interpretation', Indologica Taurinensia 12: 139-54. 1986, 'Durga on bhdva\ in E. Kahrs (ed.): Kalydnamitrdrdganam: essays in honour of Nils Simonsson, Oslo: Norwegian University Press, pp. 115-44. 1992, 'Exploring the SaddanTtV, Journal of the Pali Text Society 17: 1-112. 1996, Substitution and change: foundations of traditional Indian hermeneutics, University of Oslo. (Diss.) Kas Kdsikdvrtti of Vamana and Jayaditya, ed. with the Nydsa of Jinendrabuddhi and the Padamanjan of Haradatta by Dwarikadas Shastri and Kalikaprasad Shukla, 6 vols., Varanasi, 1965-7. KasN Jinendrabuddhi's Nydsa on the Kdsikd, in Kas. KasP Haradatta's Padamanjan on the Kdsikd, in Kas. Katre, Sadashiva L.: 1948, 'Harisvamin the commentator of the Satapatha-Brahmana: a protege of Vikramaditya the great of tradition, his date - c. 54 B.C.', Bharatiya Vidyd 9 (ShriJ&rM. Munshi Diamond Jubilee Volume, part 1) (Bharatiya Vidya BhavajvBombay): 325-40. Kielhorn, Lorenz Franz (ed.): 1868, The Paribhdsendusekhara of Ndgojibhatta; part I: the Sanskrit text and various readings (BSS 2), Bombay: The Indu Prakash Press. (References to the revised edition by K.V. Abhyankar 1962.) (tr.): 1871-4, The Paribhdsendusekhara of Ndgojibhatta, edited and explained by F. Kielhorn; part II: translation and notes (BSS 7, 9,12), Bombay: The Indu Prakash Press (7, 9) / Government Central Book Depot (12). (References to the revised edition by K.V. Abhyankar 1960.) Kiparsky, Paul: 1979, Pdnini as a variationist (PCASS, Class B, 6/Current Studies in Linguistics Series 7), ed. by S.D. Joshi, Poona-Cambridge, Mass.-London: Poona University Press and M.I.T. Press, 1982, Some theoretical problems in Pdnini's grammar (BORI Post-graduate and Research Department Series 16, 'Professor K.V. Abhyankar Memorial Lectures' (second series)), Poona: BORI. Kirste, Johann (ed.): 1859, The Unddiganasutra ofHemacandra with the author's own commentary, Wien/Bombay. KSS Kdsi Sanskrit Series, vfaranasi. KSTS Kashmir Series of Texts and Studies, published under the authority of the Government of His Highness Major (later Lieut.) General Maharaja Sir Pratap Singh Sahib Bahadur, Maharaja of Jammu and Kashmir State, by the Research Department, Jammu and Kashmir State. Kunhan Raja, C: 1936, 'The chronology of the Vedabhasyakaras', Journal of Oriental Research (Madras) 10: 256-68. 1940-1, 'Niruktavartika: a rare work till now undiscovered', Annals of Oriental Research 5: 5-16.

Abbreviations and bibliography

287

1944, The author of the Niruktavartika', Adyar Library Bulletin 8: 191. Kunjunni Raja, K.: 1964, The Niruktavdrttika of Nllakantha', Adyar Library Bulletin 28: 250-62. 1979, 'Etymology and interpretation in ancient India', in Dr. V. Raghavan Shashtyabdapurti Felicitation Volume, Madras: Kuppuswami Sastri Research Institute, pp. 180-2. KZ (Kuhns) Zeitschrift fiir vergleichende Sprachforschung auf dem Gebiete der indogermanischen Sprachen, Gottingen. Laddu, Sureshachandra Dnyaneshwar: 1967, 4A possible light on the relative age of Yaska and Patanjali', Vishveshvaranand Indological Journal 5: 58-62. Lamotte, Etienne: 1974, 'Passions and impregnations of the passions in Buddhism', in Buddhist studies in honour ofl.B. Homer, ed. by L. Cousins, A. Kunst, and K.R. Norman, Dordrecht and Boston: Reidel, pp. 91-104. Larson, Gerald James and Ram Shankar Bhattacharya (eds.): 1987, Samkhya: a dualist tradition in Indian philosophy {Encyclopedia of Indian philosophy IV), Princeton University Press. Liebich, Bruno: 1922, Materialien zum Dhdtupdtha (Sitzungsberichte der Heidelberger Akademie der Wissenschaften, Philosophisch-historische Klasse 1921: 7), Heidelberg: Winter. LSI Laghusabdendusekhara (avyayibhdvdnto bhdgah) of Sri Ndgesa Bhatta, with the Sekhara Dipaka commentary by M.M.Pt. Sri Nitydnanda Panta ParvatTya with an introduction by Pt. Gopdla Sastri Nene (KSS 27), Varanasi: Chowkhamba, 1925. 7th edn 1984. LUders, Heinrich: 1951, Varuna I: Varuna und die Wasser, aus dem NachlaB herausgegeben von Ludwig Alsdorf, Gottingen: Vandenhoeck & Ruprecht. 1959, Varuna II: Varuna und das Rta, aus dem NachlaB herausgegeben von Ludwig Alsdorf, Gottingen: Vandenhoeck & Ruprecht. Macdonell, Arthur Anthony (ed.): 1886, Kdtydyana's SarvdnukramanT of the Rgveda with extracts from Sadgurusisya's commentary entitled VeddrthadTpikd, Oxford: Clarendon Press. 1904, BD. Malamoud, Charles: 1976, Terminer le sacrifice', in Madeleine Biardeau and Charles Malamoud: Le sacrifice dansVInde ancienne (Bibliotheque de VEcole des Hautes Etudes, Section des Sciences Religieuses 79), Paris: Presses Universitaires de France, pp. 155-204. Mayrhofer Manfred Mayrhofer: Kurzgefasstes etymologisches Worterbuch des Altindischen, 4 vols., Heidelberg: Winter, 1956-80. Mbh Vydkarana-mahdbhdsya of Patanjali, ed. Franz Kielhorn, 3 vols., Bombay, 1880-5. 3rd edn, revised etc. by K.V. Abhyankar, Poona: BORI, 1962-72. MbhD MahdbhdsyadTpikd of Bhartrhari, ed. by K.V Abhyankar and V.P. Limaye (BORI Post-graduate and Research Department Series 8), Poona: BORI, 1970. MbhDms MahdbhdsyadTpikd of Bhartrhari (BORI Post-graduate"and Research Department Series), Poona: BORI, 1980. (Facsimile edn of the MahdbhdsyadTpikd manuscript.) MbhP MahdbhdsyapradTpa of Kaiyata, 5 vols., ed. by Bhargavasastrl Bhikajl JosT 1951 (vol. I,*5th edn), 1942 (vol. IV), 1945 (vol. V), Sivadatta Kudala 1912 (vol. II), and Raghunath Sarma - Sivadatta Kudala 1937 (vol. III). MbhU MahdbhdsyapradTpoddyota of Nagesabhatta, ed. as MbhP.

288

Abbreviations and bibliography

Mehendale, Madhukar Anant: 1965, 'Nirukta Notes, X: Anupa', in Dr. V.V. Mirashi Felicitation Volume, Nagpur, pp. 104-7. (Repr. in Mehendale 1978: 1-4.) 1968, Some aspects of Indo-Aryan linguistics (the Wilson Philological Lectures, 1963^-), University of Bombay. 1968a, 'Nirukta Notes, XV: Aikapadika in the Nirukta 1.14', Indian Linguistics 29: 90-6. (Repr. in Mehendale 1978: 22-30.) 1978, Nirukta notes, Series //, Pune: Deccan College Post-graduate and Research Institute. 1978a, 'Yaska's second rule of derivation', in Mehendale 1978: 57-62. 1978b, The science of etymology (niruktasastra)', in Mehendale 1978: 63-80. 1986, 'On the Brhaddevatd 2.102-103', /// 29: 117-20. Mette, Adelheid: 1973, Indische Kulturstiftungsberichte und ihr Verhaltnis zur Zeitaltersage (Akademie der Wissenschaften und der Literatur, Abhandlungen der geistes- und sozialwissenschaftlichen Klasse, 1973: 1), Mainz-Wiesbaden: Verlag der Akademie der Wissenschaften und der Literatur/Steiner. MM(P) MahdrthamahjarT of Mahesvarananda with autocommentary t (-parimala) (Trivandrum Sanskrit Series 66), ed. T. Ganapati Sastrl, Trivandrum, 1919. MNP Mahdnayaprakdsa of Rajanaka Sitikantha (in Old Kashmiri) with autocommentary (in Sanskrit) (KSTS 21), ed. Mukunda Rama Sastrl, Bombay, 1918. Mor, Mansukh Raj, (Manasukharaya Mora) (ed.): 1952, Niruktam (Nighantuh) srimanmaharsiydskdcdryyapranltam, prathamo bhdgah, sri devardjayajvakrta 'nirvacana'/ndma tikd sahitam (Nighantu with commentary by Devarajayajvan) (Gurumandala Granthamdldydh 10), Calcutta. MVUT JMdlinTvijayottaratantra (KSTS 37), ed. Madhusudana Kaul Sastrl, Srinagar, 1922. MW Monier Monier-Williams: A Sanskrit-English Dictionary (New edn, greatly enlarged and improved), Oxford: Clarendon Press, 1899. Narasimhacharya, M.S. (ed.): 1982, Mahdbhdsya PradTpa vydkhydndni: commentaires sur le Mahdbhdsya de Patanjali et le PradTpa de Kaiyata, vol. 9 (adhydya 6, pdda 1-4) (P/F/51,9), Pondichery: Institut Francais d'Indologie. NeT(U) Netratantra, with the commentary (-uddyota) of Rajanaka Ksemaraja (KSTS 46, 61), 2 vols., ed. Madhusudana Kaul Sastrl, Bombay, 1926, 1939. Nigh Nighantu. Ref. to ed. L. Sarup 1927. Nir Nirukta o$ Yaska. Ref. to ed. L. Sarup 1927. NS Nydyasutra of Gautama, ed. with Vatsyayana's Bhdsya and Visvanatha's Vrtti by Digambara Sastrl Josi (AASS 91), Poona: Anandasrama, 1922. (Ref. to Sutra or page.) NSV Niruktaslokavdrttika of Nllakantha Gargya, ed. by Vijayapalah, Calcutta: Srlmatl SavitrldevI Bagriya Trust, 1982. Ojihara, Yutaka and Louis Reitou (trs.): 1962, La Kdsikd-vrtti (adhydya I, pdda 1) traduite et commentee, 2e partie (Publications de VEcole Frangaise d'ExtremeOrient 48), Paris: Ecole Frangaise d'Extreme-Orient. Padoux, Andre: 1963, Recherches sur la symbolique et Venergie de la parole dans certains textes tantriques (Publications de VInstitut de Civilisation Indienne de rUniversite de Paris 21), Paris: de Boccard. (Revised, 2nd edn 1975.) 1990, Vac: the concept of the word in selected Hindu Tantras, Albany, N.Y.: State University of New York Press. (English translation by Jacques Gontier of revised and enlarged version of Padoux 1963.)

Abbreviations and bibliography

289

Palsule, Gajanan Balakrishna: 1955, A concordance of Sanskrit dhdtupdthas (with index of meanings) (Deccan College Dissertation Series 14), Poona: Deccan College Post-graduate and Research Institute. 1966,' "Samjnayam" in Panini', Journal of the University of Poona 25: 31-75. (Repr. as PCASS, Class A, 10, i966.) Palsule, Gajanan Balakrishna and Vamana Balakrishna Bhagavat: 1991, MahdbhdsyadTpikd of Bhartrhari; fascicule VIII: dhnika VII (BORIPost-graduate and Research Department Series 37), Poona: BORI. Pandey, Kanti Chandra: 1963, Abhinavagupta: an historical and philosophical study, 2nd edn (Chowkhamba Sanskrit Studies 1), Varanasi: Chowkhamba. Parkin, David (ed.): 1982, Semantic anthropology (Association of Social Anthropologists monographs 22), London: Academic Press. Patavardhana, Ramakrishna SastrT (Tatya SastrT), Gangadhara SastrT ManavallT, Ramacandra SastrT Kotibhaskara and GosvamI Damodara Sastrl (eds.): 1884-1937, Vdkyapadiya, a treatise on the philosophy of Sanskrit grammar by Bhartrhari, with a commentary by Punyardja (kdndas 1-2); . . . with a commentary by Heldrdja (kdnda 3) (VdkyapadTyam sdngavaiydkaranasiddhdntanirupanam sribhartrharimahdvaiydkaranaviracitam srTpunyardjakrtaprakdsdkhyatikdyutam; . . . srtheldrdjakrtaprakdsdkhyatTkdyutam) (Benares Sanskrit Series 6), 2 vols. in 11 fascicles, Benares: Braj B. Das. PCASS Publications of the Centre of Advanced Study in Sanskrit, Poona. PH Pratyabhijndhrdaya of Rajanaka Ksemaraja with autocommentary (KSTS 3), ed. J.C. Chatterji, Srinagar, 1911. PIFI Publications de Vlnstitut Frangais d'Indologie, Pondichery. PPU Unddisutras in the pahcapddT recension. Ref. to ed. T. Aufrecht 1859. PT(V) Pardtrimsikd with the commentary (-vivarana) of Abhinavagupta (KSTS 18), ed. Pt. Mukunda Rama SastrT, Bombay, 1918. pw Sanskrit-Wdrterbuch in kurzerer Fassung, bearbeitet von Otto Bohtlingk, 7 parts, St. Petersburg: Kaiserlichen Akademie der Wissenschaften, 1879-89. Quine, Willard Van Orman: 1951, Two dogmas of empiricism', Philosophical Review 60: 20-43. (Repr. in W.V. Quine 1961:20-46.) 1960, Word and object, Cambridge, Mass.: M.I.T. Press. 1961, From a logical point of view, Cambridge, Mass.: Harvard University Press, 2nd edn, revised. (1st edn 1953.) 1969, 'Ontological relativity', in W.V. Quine: Ontological relativity and other essays, New York: Columbia University Press, pp. 26-68. 1990, Pursuit of truth, Cambridge, Mass, and London: Harvard University Press. (2nd revised edition 1992.) Rajavade, Vaijanatha KasTnatha (ed.): 1921, Durgdcdryakrtavrttisametam niruktam, purvasatkdtmakah prathamo bhdgah (Yaska's Nirukta with the commentary of Durga, part 1) (AASS 88:1), Poona: Anandasrama. (ed.): 1926, SrimadydskdcdryapranTtam durgacdryakrtavrttisametam niruktam, uttarasatkdtmako dvitiyo bhdgah (Yaska's Nirukta with the commentary of Durga, part 2) (AASS 88:2), Poona: Anandasrama. 1940, Yaska's Nirukta; Volume I: Introduction. Full texts ofNighantu and Nirukta. Cursory examination ofNighantu. Notes on chapters I-III of Nirukta. 25 Indexes (Government Oriental Series, Class A, 7), Poona: BORI. Rau, Wilhelm (ed.): 1977, Bhartrharis Vdkyapadiya: die Mulakdrikds nach den

290

Abbreviations and bibliography

« »

Handschnften herausgegeben und mit einem Pdda-Index versehen {Abhandlungen fur die Kunde des Morgenlandes XLII,4), Wiesbaden: Steiner. Renou, Louis: 1941-2, 'Les connexions entre le rituel et la grammaire en Sanskrit', JA 233: 105-65. (Repr. in J.F. Staal (ed.): A reader on the Sanskrit grammarians {Studies in Linguistics 1), Cambridge, Mass.: M.I.T. Press, 1972, pp. 435-69.) 1942, Terminologie grammaticale du Sanskrit, 3 vols. {Bibliotheque de VEcole des Hautes Etudes, fascs. 280-2), Paris: Champion. (Repr. in one vol., Paris: Champion, 1957.) 1955, 'Les nipdtana-sutra de Panini et questions diverses', EVP I: 103-30. Renou, Louis and Jean Filliozat: 1953, Ulnde classique; manuel des etudes indiennes; tome II, Paris: Ecole Francaise d'Extreme-Orient. Rocher, Rosane: 1968, La theorie des voix du verbe dans Vecole pdnineenne (le I4e dhnika) (Universite Libre de Bruxelles, Travaux de la Faculte de Philosophie et Lettres 35), Bruxelles: Presses Universitaires de Bruxelles. Roth, Rudolph (ed.): 1852, Jdska's Nirukta sammt den Nighantavas, herausgegeben und erldutert von Rudolph Roth, Gottingen: Dieterichschen Buchhandlung. Ruegg, David Seyfort: 1958, 'On the term buddhivipanndma and the problem of illusory change', /// 2: 271-83. RV Rgveda. Ref. to Rgveda-samhitd srTmatsdyandcdrya-viracita-bhdsyasametd {Rgveda-samhitd with the commentary of Sayana), ed. N.S. Sontakke et al., 5 vols., Poona: Vaidika Samsodhana Mandala, 1933-51. (Vol. 1 repr. 1972.) Sanderson, Alexis: 1985, 'Purity and power among the Brahmans of Kashmir', in Michael Carrithers, Steven Collins, and Steven Lukes (eds.): The category of the person, Cambridge University Press, pp. 190-216. 1986, 'Mandala and Agamic identity in the Trika of Kashmir', in Mantras et diagrammes rituels dans Vhindouisme (Table ronde: CNRS, Equipe n° 249 'L'hindouisme: textes, • doctrines, pratiques'), Paris: Editions du CNRS, pp. 169-214. 1988, 'Saivism and the Tantric traditions', in Stewart Sutherland et al. (eds.): The world's religions, London: Routledge, pp. 660-704. (Repr. in Friedhelm Hardy (ed.): The world's religions: the religions of Asia, London: Routledge, 1990, pp. 128-72. All references are to this edition.) 1990, 'The visualization of the deities of the Trika', in Andre Padoux (ed.): Limage divine: culte et meditation dans Vhindouisme (CNRS, UPR 249 'L'hindouisme. textes, doctrines, pratiques'), Paris: Editions du CNRS, pp. 31-88. 1992, 'The doctrine of the Malinlvijayottaratantra', in Teun Goudriaan (ed.): Ritual and speculation in early Tantrism: studies in honor of Andre Padoux, Albany, N.Y.: State University of New York Press, pp. 281-312. 1995, 'Meaning in Tantric ritual', in Anne-Marie Blondeau and Kristofer Schipper (eds.): Essais sur le rituHJII (Colloque du centenaire de la Section des Sciences Religieuses de VEcole Pratique des Hautes Etudes) {Bibliotheque de VEcole des Hautes Etudes: Section des Sciences Religieuses CII), Lou vain-Paris: Peeters, pp. 15-95. Samp, Lakshman (ed., tr.): 1920-7, The Nighantu and the Nirukta, the oldest Indian treatise on etymology, philology, and semantics. (Repr. Delhi: Motilal Banarsidass, 1967.) 1920, {id.) Introduction, Oxford University Press. 1921, {id.) English translation, Oxford University Press.

Abbreviations and bibliography

291

1927, (id.) Sanskrit text, with an appendix showing the relation of the Nirukta with other Sanskrit works, Lahore: University of the Panjab. (ed.): 1928, Fragments of the commentaries of Skandasvdmin and Mahesvara on the Nirukta, Lahore: University of the Panjab. (Contains the text of chapter one.) 1929, Indices and appendices to the Nirukta with an introduction, Lahore: University of the Panjab. (ed.): 1931, Commentary of Skandasvdmin and Mahesvara on the Nirukta: chapters II-VI, Lahore: University of the Panjab. (ed.): 1934, Commentary of Skandasvdmin and Mahesvara on the Nirukta: vols. Ill & IV, chapters VII-XIII, Lahore: University of the Panjab. 1937, 'Date of Skandasvamin', in Gangdndtha Jhd commemoration volume, ed. by K. Chattopadhyaya et al (Poona Oriental Series 39), Poona: Oriental Book Agency. Sarveswara Sharma, Peri: 1972, The Kdlasamuddesa of BhartrharVs VdkyapadTya (together with Heldrdja's commentary translated from the Sanskrit for the first time), Delhi: Motilal Banarsidass. SatBr Satapathabrdhmana in the Madhyandina recension, ed. A. Weber, Berlin, 1849-55. (2nd repr. (Chowkhamba Sanskrit Series 96), Varanasi: Chowkhamba, 1964.) Say Sayana's introduction to his commentary on the Rgveda. Ref. to page in N.S. Sontakke et al. (eds.): Rgveda-samhitd srfmatsdyandcdrya-viracita-bhdsyasametd (Rgveda-samhitd with the commentary of Sayana), vol. 1, Poona: Vaidika Samsodhana Mandala, 1933. (Repr. 1972.) Scharfe, Hartmut: 1977, Grammatical literature (HIL V,2), Wiesbaden: Harrassowitz. Schneider, Ulrich: 1954, Acht Etymologien aus dem Aggaiina-Sutta', in Asiatica: Festschrift Fnedrich Wellerzum 65. Geburtstag, ed. J. Schubert and U. Schneider, Harrassowitz: Leipzig, pp. 575-83. Simonsson, Nils: 1957, Indo-tibetische Studien: die Methoden der tibetischen Ubersetzer, untersucht im Hinblick auf die Bedeutung ihrer Ubersetzungen fur die Sanskritphilologie, Uppsala: Almqvist & Wiksell. 1982, 'Reflections on the grammatical tradition in Tibet', in Indological and Buddhist studies: volume in honour of Professor J. W. de Jong on his sixtieth birthday, ed. by L.A. Hercus et al., Canberra: Faculty of Asian Studies, pp. 531-44. Simson, Georg von: 1988, 'Etymologie als Mittel ideologischer Auseinandersetzung: Bemerkungen zum Aggafinasutta des Dighanikaya', in Studia Indogermanica et Slavica: Festgabefur Werner Thomas zum 65. Geburtstag (Specimina Philologiae Slavicae, Supplementband 26), ed. by P. Kosta et al., Munich: Otto Sagner, pp. 87-98. Singh, Fatah: 1962, The Vedic etymology: a critical evaluation of the science of etymology as found in Vedic literature, Kota: The Sanskriti-Sadan. Singh, Maan: 1994, The Upanisadic etymologies, Delhi: Nirmal Publications. Sltaramacari Sastri (ed.): 1926/7 (Samvat 1983), VdkyapadTyabrahmakdndah yogirdjamahdvaiydkaranasnbhartrhariviracitah . . . srldravyesajhdsdstripratyekdrthaprakdsikdfikdparigumphitah . . . snsitdrdmdcdrisdstrind samsodhitah (VdkyapadTya, brahmakdnda, with the commentary Pratyekdrthaprakdsikd of Dravyesa Jha Sarma), Vrindavan: Brajendra Press. Sivanarayana Sastri: 1969, Nirukta-mTmdmsd, Delhi-Varanasi: Indological Book House.

292

Abbreviations and bibliography

1972, Nirukta ke pamca adhydya (1-4 tathd 7 adhydya) rju-vivrti ndmaka vydkhyd sameta, Delhi-Varanasi: Indological Book House. SK Sabdakaustubha of Bhattoji Dlksita, (Sabdakaustubhah snmadbhattojidTksitaviracitah, prathamddhydyasya prathamapddo navdhnikaparyantah), vol. 1, fasc. \-A, ed. Gopala Sastri Nene and Mukunda Sastrl Puntamkar, Benares: Chowkhamba, 1933. SK Siddhdntakaumudi of Bhattoji Dlksita, ed. with the Bdlamanoramd of Yasudeva Dlksita and the TattvabodhinT of JnanendrasarasvatI by Giridhara Sarma and Paramesvarananda Sarma , 4 vols., Varanasi, 1958-61. Skold, Hannes: 1926, The Nirukta, its place in old Indian literature, its etymologies (Acta Regiae Societatis Humaniorum Litterarum Lundensis 8), Lund: Gleerup. SM The commentary of Skandasvamin and Mahesvara on the Nirukta', 4 vols.: vol. I Sarup 1928, vol. II Sarup 1931, vols. Ill & IV Sarup 1934. (Ref. to vol., page and line.) Smith, Helmer (ed.): 1928-30, SaddanTti: la grammaire palie dAggavamsa: I Padamdld, II Dhdtumdld, III Suttamdld, Lund: Gleerup. 1949-66, id.: IV-V,2 Tables, Lund: Gleerup. SN Spandanirnaya, in: The Spandakdrikds of Vasugupta with the Nirnaya by Ksemardja, edited with preface, introduction and English translation by Madhusudan Kaul Sastrl (KSTS 42), Srinagar, 1925. Snellgrove, David: 1959, The Hevajratantra: a critical study, part 1: introduction and translation {London Oriental Series 6), London: Oxford University Press. Speijer, J.S.: 1886, Sanskrit syntax, Leyden: Brill. (Repr. Delhi: Motilal Banarsidass, 1973.) SSS Sdnkhdyana Srautasutra, ed. with the commentary of Varadattasuta Anartlya by Alfred Hillebrandt, 4 vols. (Bibliotheca Indica 99), Calcutta: Asiatic Society of Bengal, 1885-99. SS(V) Sivasutra with the commentary (-vimarsini) of Rajanaka Ksemaraja (KSTS 1), ed. J.C. Chatterji, Srinagar, 1911. Staal, Johan Frederik (Frits): 1967, Word order in Sanskrit and universal grammar (Foundations of Language, Supplementary Series 5), Dordrecht: Reidel. Stein, Marc Aurel: 1900, Kalhana's RdjataranginT, a chronicle of the kings ofKasmir. Translated with an introduction, commentary and appendices, 2 vols., Westminster: Archibald Constable. Subramania Iyer, K.A. (ed.): 1966, Vdkyapadiya of Bhartrhari with the Vrtti and the Paddhati of Vrsabhadeva, kdnda I (Deccan College Monograph Series 32), Poona: Deccan College. 1983, Vdkyapadiya of Bhartrhari (an ancient treatise on the philosophy of Sanskrit grammar) containing the Tikd ofPunyardja and the ancient Vrtti, kdnda 2, Delhi: Motilal Banarsidass. v SvT(U) Svacchandatantra with the commentary (-uddyota) of Rajanaka Ksemaraja (KSTS 31, 38, 44, 48, 51, 53, 56), 6 vols., ed. Madhusudan Kaul SastrT, Bombay, 1921-35. TA(V) Tantrdloka of Abhinavagupta with the commentary (-viveka) of Rajanaka Jayaratha (KSTS 23, 28, 30, 36, 35, 29, 41, 47, 59, 52, 57, 58), 12 vols., ed. Mukunda Rama Sastrl (vol. 1) and Madhusudan Kaul Sastrl (vols. 2-12), Bombay and Srinagar, 1918-38.

Abbreviations and bibliography

293

Taranatha Tarkavacaspati: 1873-84, Vdcaspatyam (brhat samskrtdbhidhdnam)... sri tdrdndthatarkavdcaspatibhattdcdryena sankalitam (Vachaspatyam (a comprehensive Sanskrit dictionary) compiled by Sri Taranatha Tarkavacaspati), 6 vols., Calcutta. (Repr. in 6 vols. (Chowkhamba Sanskrit Series 94), Varanasi: Chowkhamba, 1962.) Tarski, Alfred: 1935, 'Der Wahrheitsbegriff in den formalisierten Sprachen', Studia Philosophica 1: 261-405. (German translation of a book in Polish, 1933.) 1956, 'The concept of truth in formalized languages', in Logic, semantics, metamathematics, Oxford: Clarendon Press, pp. 152-278. (Collected papers 1923-38.) Thieme, Paul: 1935, 'Zur Datierung des Panini', Zeitschrift der deutschen Morgenldndischen Gesellschaft 89: 21-4. (Repr. Buddruss 1971:528-31.) 1955, review of Jacob Wackernagel, Altindische Grammatik II2: Albert Delbrunner: Die Nominalsuffixe, Gottingen, 1954, in Gottingische Gelehrte Anzeigen 209: 182-216. (Repr. Buddruss 1971:661-95.) 1957, 'The interpretation of the learned', in Felicitation volume presented to Professor Sripad Krishna Belvalkar, ed. by S. Radhakrishnan, V.V. Mirashi, R.N. Dandekar, S.K. De, V. Raghavan, and A.S. Altekar, Varanasi: Motilal Banarsidass, pp. 47-62. (Repr. Buddruss 1971:596-611.) 1958, review of 1957 repr. of Renou 1942, in Gottingische Gelehrte Anzeigen 212: 19-49. (Repr. Buddruss 1971:727-57.) 1961, Introduction to Betty Shefts: Grammatical method in Panini: his treatment of Sanskrit present stems (American Oriental Series Essay 1), New Haven: American Oriental Society, pp. ix-x. 1968, 'Adesd1 in Melanges d'indianisme a la memoire de Louis Renou (Publications de Vlnstitutde Civilisation Indienne, serie in-8° 28), Paris: Boccard, pp. 715-23. (Repr. Buddruss 1971:259-67.) Turner, Ralph L.: 1979, review of M.A. Mehendale 1978, Indian Linguistics 40: 212. van Nooten, Barend A.: 1970, 'Sanskrit gamsyate, an anit future', JAOS 90: 159. Varma, Siddheshwar: 1953, The etymologies ofYdska, with the assistance of Bhim Dev Sastrl (Vishveshvaranand Indological Series 5), Hoshiarpur: Vishveshvaranand Vedic Research Institute. VB Vedic Bibliography by R.N. Dandekar, vol. 1: 1946 (repr. 1986), vol. 2: 1961 (repr. 1978), vol. 3: 1973, vol. 4: 1985 (Government Oriental Series, Class B, 10,11, 13, 14), Poona: BORI. VBh Vijndnabhairavatantra with the commentary (Uddyota) of Sivopadhyaya on vv. 24 to the end incorporating the surviving fragment of the commentary (Vivrti) of Rajanaka Ksemaraja on vv. 1 to 23 (KSTS 8), ed. Mukunda Rama Sastrl, Bombay, 1918. VBhK Vijndnabhairavatantra with the commentary (-kaumudt) of Ananda Bhatta (KSTS 9), ed. Mukunda Rama Sastri, Bombay, 1918. VBhU Vijndnabhairavoddyota of Rajanaka Ksemaraja; in VBh. VBhV Vijndnabhairavavivrti of Sivopadhyaya; in VBh. Vedavrata (ed.): 1962-3, SrTbhagavatpatahjaliviracitam vydkarana-mahdbhdsyam (srTkaiyyatakrtapradipena ndgojibhattakrtena bhdsyapradipoddyotena ca vibhusitam) (Patanjali's Mahdbhdsya with Kaiyata's Pradipa and Nagesa's Uddyota), 5 vols., Gurukula Jhajjar (Rohatak): Haryana Sahitya Samsthana. Verma, Nargis: 1991, The etymologies in the Satapatha Brdhmana, Delhi: Nag Publishers.

294

Abbreviations and bibliography

Vijayapala (ed.) {sampddakah vijayapdlah): 1982, Nirukta-slokavarttikam, Calcutta: Srlmatl SavitrldevI Bagriya Trust (distributed and printed by the Ram Lai Kapur Trust, Bahalgarh, Sonlpat), {upodghdtah pp. 1-52, text pp. 1-588). VP VdkyapadTya of Bhartrhari, ed. Wilhelm Rau, Wiesbaden: Steiner, 1977. vt. vdrttika. Wezler, Albrecht: 1972, 'Marginalien zu Paninis Astadhyayl I: sthdnin\ KZ 86: 7-20. 1994, 'Once again on Patanjali's definition of a word (Studies inPatanjali's Mahabhasya V)', Wiener Zeitschrift fur die Kunde Sudasiens 38: 173T89. Whitney, William Dwight: 1871, The Tdittiriya-Prdtigdkhya, with its commentary, the Tribhdshyaratna: text, translation, and notes, New Haven: for the American Oriental Society. (From the Journal of the American Oriental Society 9, 1871.) YD YuktidTpikd, ed. by Ram Chandra Pandeya, Delhi: Motilal Banarsidass, 1967. Yudhisthira Mlmamsaka (ed.): 1943, Dasapddyunddivrtti {Princess of Wales Sarasvati Bhavana Texts Series 81), Benares: Government Sanskrit College. 1965/6 (Samvat 2022), Niruktasamuccayah dcdryavararucipramtah (Vararuci's Niruktasamuccaya), Ajmer: Bharatiya Pracyavidya Pratisthana (2nd edn). 1965/6a (Samvat 2022), Kdsakrtsna-dhdtuvydkhydnam, sricannavtrakavikrtakarndtakatlkdydh samskrtarupdntaram {Kdsakrtsna-dhdtuvydkhydnam with Sanskrit translation of the Kannada commentary by Cannavlrakavi), Ajmer: Bharatiya Pracyavidya Pratisthana. 1973, Samskrta vydkarana-sdstra kd itihdsa (tin bhdgom mem puma), prathama bhdga, Sonipat: Ram Lai Kapur Trust, 3rd edn Samvat 2030.

I N D E X OF T E X T - P L A C E S

Atharvaveda 3.13.2 26, 53, 103 Rgveda (RV) 1.164.33 143; 4.58.3 186; 7.75.5 Abhidharmakosabhdsya (AKBh) 270,5-6 275 25-6 ' \ AstddhyayT(A) 1.1.1 165; 1.1.47 257; 1.1.49 ' 177, 189-219, 229-36, 236, 237, 238, 239, Kdsikdvrtti (Kas) 11:209-11 237 n 152; 11:284 256n.201;V:398 257 240, 241, 244, 245, 247, 249, 256, 257, 265, 266; 1.1.52 202, 220, 223, 231, 234; 1.1.54 220, 221, 225, 231, 234; 1.1.55 220, 225, 234; Chdyd of Vaidyanatha Payagunda (ed. BhargavasastrT JosI 1951) 412, note 4 216; 1.1.56 188, 216, 269; 1.1.57 264 n.213; 1.1.60 412, note 10 233 163, 193 n.34; 1.1.61 193 n.34; 1.1.62 275; 1.1.68 163; 1.1.69 222; 1.1.72 220, 224, 227, 230, 234; 1.2.45 193 n.33; 1.2.46 193; 1.3.11 Jainendramahdvrtti (JV) 1.1.46 262 210; 1.4.13 201 n.60, 234; 1.4.14 274; 1.4.17 224 n.117; 1.4.18 224 n.117; 1.4.51 276; 2.1.1 Tantrdloka (TA) 1.95 63-6; 1.96 72-4; 1.97 243; 2.2 24 193; 2.3.50 237; 2.4.52 189, 75-6; 1.98 77; 1.99-100ab 78; lOOcd 83; 237, 238, 240, 252, 253, 256, 257, 258, 261, 1.235-6 85; 1.237-40 93-4, 3.65 73 n.48, 265, 266; 2.4.71 193; 3.1.97 256n.201; 5.135d-6ab 71; 6.30-lab 67; 6.31cd-33 3.1.124 256 n.201; 3.2.110 51, 175 n.3; 68-9; 6.228c-30b 275 3.2.174 74; 3.3.18 195; 3.3.19 195, 3.3.94 Tantrdlokaviveka (TAV) 1[1] 139,7-8 63; 195; 3.3.115 53, 248; 3.3.117 53, 249; 1[1]139,13-14 65; 1[1] 141,1-4 72; 3.3.122 50; 3.4.69 51, 175 n.3; 3.4.113 189 l[l]141,4-5 73; 1[1] 141,5-7 74; n.31; 3.4.114 189 n.31; 4.1.82 242; 4.1.83 75; 1[1]141,8~14 76; 1[1] 141,14-142,4 77-8; 4.3.120 75; 5.2.94 206, 242, 243, 244; l[l]142,4-10 78-9; l[l]142,10-ll 81; 5.2 127 206; 5.3.71 225; 6.1.63 232; 6.1.77 1[1]143,1-11 83-4; 2[3]74,12 79 n.59; 252 n.192, 253, 263, 264, 266; 6.3.1 193; 4[6]30,ll-31,8 67; 4[6]31,9-32,2 67-8; 6.3.9 193; 6.3.32 185; 6.4.1 212,245,246, 4[6]32,3-33,8 68-9 247; 6.4.2 212; 6.4.24 202, 247; 6.4.34 201, Taittinya Prdtisdkhya 1.56 274 202, 203, 209, 247; 6.4.35 209, 247; 6.4.47 257, 258, 263; 6.4.77 222; 6.4.85 222; 6.4.89 Durga's commentary on the Nirukta (D) 201, 202, 203; 6.4.130 224, 225, 230, 231, 1:30,1-4 29 n.42; 1:30,2-3 161; 235; 7.1.9 245; 7.2.86 227; 7.2.91 227; 1:30,3-4 46; 1:77,10-11 159; 1:88,4-6 110; 7.2.93 227; 7.2.102 226; 7.2.106 226; 7.3.43 1:154,18-19 35; 1:169,8-10 114; 1:179,7 115; 252 n.192; 7.3.103 193; 7.4.27 222n.ll3; 1:180,8-11 116; 1:191,1-2 134; 7.4.28 222 n.113; 7.4.32 221; 7.4.53 231 1:207,15-208,3 172; 1:226,4-5 160; 1:229,1-2 n.139; 8.1.62 276; 8.1.63 276; 8.2.83 185; 136; 1:281,14-17 137; 1:292.6-7 124; 8.4.61 220, 225; 8.4.65 220 n.108 1:318,5-6 119; 1:318,13 1J9; 1:364,2-4 139; 1:417,1 142; 1:445,9-10 143; 1:479,17-18 Apastamba Srautasutra (ASS) 24.3.52-4 188 145; 1-527,8-10 146; 1:566,12 127, Asvaldyana Grhyasutra (AGS) 1.7.1-2 184 1:566,13-14 128; 1:575,15-18 147; Asvaldyana Srautasutra (ASS) 1.1.2-3 184; 1:581,17-582,2 128; 1:650,15-16 148; 2.1.1 184 1:812,5-6 130; 11:924,3 99; 11:931,11-13 99 n.7; 11:970,9-10 121; 11:981,12-13 117; Uddyotana (ed. Narasimhacharya 1982) 269 11:103947-18 156; 11-1085,12-14 150 248 Desopadesa (Des) S3 94-5

Index of text-places Nirukta (Nir) 1.4 105, 106, 161; 1.6 106, 159; 256 n.201; 11:391,17-23 242; 111:179,7-8 1.7 106, 107, 108, 169; 1.8 108, 109; 1.9 109, 246; 111:222,16-17 219 110, 111; 1.12 35; 1.15 31, 32; 1.17 112, 154; MahdbhdsyadTpikd (MbhD) 288,10-12 195; 1.18 46; 1.19 46; 1.20 28, 46, 132; 2.1 36, 288,17-289,2 239^0; 289,2-9 251; 291,1-3 37, 38, 91, 92; 2.2 32, 100, 114; 2.3 38; 2.5 206; 293,10-23 229-30 115, 116, 132, 152; 2.6 133; 2.7 38; 2.8 134; Mahdbhdsyapradipa (MbhP) I:46a 120; I:56b 2.13 134; 2.15 101, 154; 2.16 27; 2.17 135; 42; I:62a 250; I:62b 43; I:64b 45; I:408a 2.18 26, 159, 171; 2.20 54, 104; 2.21 158; 194; I:408b 239, 255-6; I:409a 197; I:409b 2.22 136; 3.1 159; 3.9 137; 3.10 137; 3.11 202, 204; I:410a 205;I:411a 211, 237 n.150; 118; 3.13 123, 158; 3.18 118; 3.19 103; 3.20 I:411b 212;I:412a 215; I:412b 231;I:453a 125; 4.1 32; 4.5 138; 4.8 139; 4.10 120; 4.13 271;I:453b 272; I:454a 261, 273; II:318b 140; 4.16 141, 159; 4.18 142; 4.19 142; 4.21 243; IV:343b-344a 243; V:270a 247; V:354a 142; 4.25 126; 4.27 54, 104, 143; 5.1 143; 225 5.2 144; 5.4 144; 5.5 54,103; 5.7 145; 5.14 Mahdbhdsyapradipoddyota (MbhU) 1:64b 45; 145; 5.22 146; 5.26 127, 147; 6.1 147; 6.17 I:65b 199;I:408a 195; I:408b 200,241; 148; 6.21 155; 6.22 117, 155; 6.27 155; 6.30 I:408b-9a 258; I:409a 197; I:409b 202, 204; 120; 7.9 129; 7.13 29 n.40; 7.15 100; 7.20 I:410a 205, 207, 237 n.151; I:411b 212, 162; 7.23 130; 7.29 117, 149; 9.8 149; 9.11 I:412a 216; I:412b 231-2; I:453b 271;I:454a 121; 9.14 130; 9.22 100; 9.26 27,53,98; 260-1; I:454b 261-2; IV:343b 243; V:270b 10.10 121; 10.17 117, 156; 10.22 98; 10.31 246, 247 122; 10.44 149; 11.5 122, 156, 158; 11.6 Mahdrthamanjariparimala (MMP) 4,10 94; 159; 11.24 131; 11.39 150; 11.46 171; 11.47 179,25-6 88 151; 12.5 171; 12.7 162, 177; 12.26 122; 12.40 162 Laghusabdendusekhara (LSI) 58,1-2 244; 58,6 Niruktaslokavdrttika (NSV) 1.6.28 114; 1.6.151 248-9; 58,6-9,2 263; 59,2-6 264 132 VdkyapadTya (VP) 2.482 10; 1.49-50 254-5 Nydyasutra (NS) 1.1.1 198; 1.1.24 198 Vijndnabhairavatantra (VBh) 130 71, 90 Nydsa(K2isN) 1:168 248,249 Vijndnabhairavavivrti (VBhV) 114,7-9 89; 114,9-10 90; 114,14 90; 115,10-13 90 PadamanjarT (Kas?) 1:169 259 Vijhdnabhairavoddyota (VBhU) 1,2-5 61-2, Pardtnmsikdvivarana (PTV) 99,9 92 86-7 Paribhdsendusekhara (ed. Abhyankar 1962) 28,11-12 228 Satapathabrdhmana 6.1.1.9 27, 53 Pratyabhijfidhrdaya (PH) 28,12-29,4 80 Sabdakaustubha (SK) 244,10-12 216; n.63 ^ 244,12-15 218 Sdnkhdyana Srautasutra (SSS) 1.16.1 184 Bdlamanoramd (SK) 1:44,4-5 214, 242 Brhaddranyaka Upamsad (BAUp) 2.3.11 [M]Sivasutra 3.29 92 SivasutravimarsinT (SSV) 116,7-118,4 92-3 ' 2.3.6 [K] 181 Bodhapancadasikd (BPD) 4 73; lied 83 Sayana's introduction to his commentary on the Baudhdyana Srautasutra (BSS) 24.8 182 Rgveda (Say) p. 28 29, 30, 31, 47; p. 29 33, Bhdradvdja Srautasutra (BhSS) 5.17.1-3 184; 46, 47, 49 6.15.4-5 184; 6.15.7 188; 14.1.1-3 184 Skanda-Mahesvara's commentary on the Nirukta (SM) 1:46,6-8 50; 1:51,11 105; Mahdbhdsya (Mbh) 1:4,12 120; 1:6,8-11 41-2; 1:67,7-10 160; 1:75,8 110; 1:75,11 110; 1:7,8-10 43; 1:7,11-12 44; 1:7,23-5 44; 1:104,14 113; 1:104,14-15 170; 11:42,12 115; 1:29,21-4 223; 1:40,20-1 165; 1:40,26-8 11:45,3 117; 11:45,5-6 170; 11:63,5 134; 165; 1:75,13-14 187; 1:118,6=^192, 236; 11:90,3-4 101; 11:90,4-6 102; II: 149,7-9 1:118,8-13 197; 1:118,13-17 201; 1:118,18-25 138; 11:152,17 118; 11:156,11-13 124; 203; 1:119,1 204; 1:119,2-3 206; 1:119,4-15 11:183,10 118; 11:183,14 119; 11:270,18-20 208; 1:119,16-19 211; 1:119,19-20 212; 127; 11:281,18-19 143; II:322,12T13 145; 1:119,20-5 213; 1:119,25-8 218; 1:126,15-17 11:425,14 130; 11:482,14-15 156; 111:89,1-2 249; 1:137,4-6 270; 1:137,8-11 271; 170; 111:156,1 130; IV: 101,14-201,1 172 1-137,13-14 272; 1:137,13-18 260; Spandanirnaya (SN) 38,9 80 1:137,17-26 272; 1:137,18-26 261; 1:158,2-8 Svacchandatantwddyota (SvTU) l[l]:3,12-4,7 163; 1:217,22-3 237; 1:324,7-10 49; 87-8; 3[5]:76,17-18 94; 6[15]: 146,11-16 1:363,24-8 47; 11:31,10 256 n.201; H: 199,21 60-1

I N D E X OF WORDS ANALYSED BY THE NIRVACANA METHOD

amhatih 126-7 amhah 126-7 amhurah 155 amhuh 121 amhuranam 155 akupdrah 142 aksi 33, 109 agham 129 anuh 117 adhvd 67-9 anupah 136 anupdh 136 andhah 143 «n/i 145 ardham 125-6 104 135 131 6, 53, 100, 103 dsyam 111 o/w/wz/i 144 dhutih25 indrah 25 wAa 149 z: isikd 149 116, 169 usdh 26, 159, 171, 172 rjuh 155 kaksah 114 kacchapah 142 kacchah 142 fcarraz/z'llO 159 154 154 155 z 118 krttih 146

krstayah 98 toia/i 147 kruram 155 ksTram 116 137 gdyatram 108 132 158 gaurah 150 gauri 150 gau/t 115, 132, 152, 162 grismah 54, 104 candrah 156, 158 122 106 142 talam 127 ftz/u 127 154 daksinah 107 daksind 107 ' 100 130 difafi 93-4, 95 devah 70 n.43, 100 dyuh 159 (i/zanva 54, 103 dhuh 137 142 ndma 143 «f/?«M/i 49-50 niruktam 47 parjanyah 121 puskaram 145 prthivT 25

298

Index of words analysed by the nirvacana methoc vikatah 120 vwfyitf'99 vivaksase 123 visam 122 v*/z 133 vrra/i 108 vrsabhah 100 vratam 134 sakvaryah 109 j/ra/i 140'
^MCI/I 147

bandhuh 142 budhnam 149 bhagah 107 bhimah 132 bhismah 132 bhairavah 61-93, 95-7, 72, 73, 74, 75, 76, 77, 78, 82, 84, 86, 87, 88, 89 magham 106 madhu 122, 139 mithunau 117, 149 musah 138 musikdh 138 mrtyuh 159 meg/ia/i 158, 161, 162, 277 yora'/z 134 rathah 121 laksanam 214 toj 127 vaksah 141 vaykfe 105, 162 vardhah 144 vavaksitha 123 v<zsf 159

sundhyuh 141 seva/i 117, 156 iva 118 safcta/i 120 5an^i/2 130 sararukam 128 salalukam 128 simhah 118 sindhuh 98-9 suvidatram 129 srprah 148 skandhah 148 sthdnam 250 /ian/* 142 /wwtaft 107

GENERAL INDEX

A 1.1.49 177, 189-219, 229-36, 236, 237, 238, artha 39-47, 164, 198, 253, 279; as meaning of sthdna 217, 235, 238, 250, 252, 253, 254, 255, 239, 240, 241, 244, 245, 247, 249, 256, 257, 256, 258, 263, 264, 265, 266, 267, 269, 277, 265, 266; basic meaning of 196, 221, 278, indigenous discussions of 41-7, Louis concluding remarks on discussion of 233-6, nirdisyamdna-panbhdsd and 218, 219-29, Renou on 40, Paul Thieme on 40, semantic range of 164 231, 234, purpose of 197-219, relation to A 1.1.52, 54, 55 and 56 234, survey of arthavdda 7, 274 Patanjali's remarks on 190-2, vdrttikas on Astddhydyl 10, 13, passim; the rules of 185 191, 197, 200, 201, 203, 208, 209, 234 atiparoksavrtti 35, 37 Abhayanandin 262 avyutpanna 32 Abhinavagupta 58, passim; on nirvacana ayogavdha sounds 223 analysis and reliance on tradition 66-71, task in exegesis 59 bahiranga 215 abhydsa 124 bandhu 25, 180 ablative case: attributes to -tehl-eh forms beliefs 3, 4, 5, 7, 62, 63, 93, 95, 277 123-31, -tehl-eh forms and'105, 169-74, 278 Bhairava 57-61; autonomous agent 63, cause of adesa 176, 178-83, 233, 238, 269, 270, 271, 274, the fear-cry 75-6, descriptive name 48, 278; in Brahmana equations 278, m ritual desiccator of time 77-8, favouring those texts 182-3, mark added to 211, Paul Thieme terrified by samsdra 74-5, goddesses of on 178-82 cognition and 78-81, &ara£a-analysis of 82-3, adhikdra 208, 209-10 Lord of the Wheel of Powers 81, master of the adhikdra-words 31, 101, 102, 103; rasmindmdni Bhlravas 79, most terrible 81-2, nirvacanas 101-2, stotrnamani 102-3 of 61-93, 95-7, rava 'the roar' 73-4, 75-6, adhvan 67-9 78-9, subjective awareness ipardmarsa) 75-6, adjacency, see dnantarya supporter of the universe 72-4, supported by ddyantaviparyaya 115 the universe 72-4, unconditioned selfAghoresvarl58 representation 75, 85, vaiydkarana derivation Agnihotram Ramanuja Tatachariar 32 of 74-5 Ahirbudhnyasamhitd 57 Bhairavatantras 58, 59 aikapadika 38 n.57 Bhairavl 58 Aitihasikas 27, 273 Bhartrhari 254; on sthdna 250-5 akAC 225-6, 228 bhdsd 185 dnantarya 192, 213, 214, 217, 235, 241, 244 bhdsyakdra 131 -andt 98-103,168, 173; causal explanation 99, Bhattoji Dlksita 196 n.46 103, 168 Bhucarl 78-81 Annambhatta 248 Bhutatantras 58 antaranga 215, 217, 218, 221, 235, 268 Brahmanas 25, 180, 181, 278 anvartha 47-50, 63-6, 67, 68, 70, 83, 84, 85, 96 Brhaddevatd 18, 19, 20, 21 anvarthasamjnd 48-50 Brhaspati 70, 71, 88 anvaya 200 apakarsa 249, 259, 262, 265 chandas 185 Apara 59 change 278 apavdda 185 commentaries, the role of 10-11

300

General index jdtipaksa 222 n.U2 Jayaratha 63 n.23, passim JnanendrasarasvatI 196 n.48 jndpaka 216 n.98, 234

Kaiyata 41, passim KalasamkarsinI 60, 72 ' Kali 58^59, 60,78, 80 Kapalika 58, 78 kdraka 49, 50-4, 103, 159-60, 168-9; analysis of bhairava 82-3 ekapada 38 n.57 kdrya 165, 186, 278 ellipsis 276 Kashmir 55, 56, 57 kasmdt 99, 131, 151, 170 equations 25, 278 etasmddeva 99, 131-52, 173; causal explanation Katyayana 13 Kaula 59, 78 151, 152, picking up kasmdt 151 KhecarT 78-81 eternality, see nityatva Krama 58, 59, 78; goddesses of cognition 78-81, etymology 24, 269; popular 56 exegesis: tradition in 57, 88, Abhinavagupta and permeating the Trika 78 Ksemaraja 60, 61; mangalasloka by 61-3, 86-89 59, potential for 91 Ksemendra 93, 94 Frege, Gottlob 26, 164 lakdras 51-2, 175 laksydpeksana 215, 221 genitive 197, 198, 268, 269, 277, 278; attributes to -tehl-eh forms 112-23, defined by rule A language, Indian grammarians' model of 188 1.1.49 190-219, 229-236, 236, 238, 240, 241, lopa 163 indicating element to undergo substitution 176, interpreted contextually 192, 202, 226, Mahdbhdsya 10, 13, passim 227, 228, 233, 247, 278, marks relation mdhdtmya texts 55-6 between word and that which it signifies 152, Mahdvyutpatti 1 153, meanings of 197-200, nirdisyamdna- Mahesvara, see Skanda-Mahesvara \ paribhdsd and 219-29, of satah 160, 162, 167, Mahesvarananda 88 277, partitive etc. 201, 202, 203, -204, 226, Mdliriivijayottaratantra 59; Abhinavagupta's 227, 246, 248, relation marked by 200, 201, exegesis of 59, ritual and metaphysics of 60 205, 213, 214, 215, 268, 269, substitutionai, mantras 57, 58, 73, 74; varnasdrupya and 92 see sthdnasasthT, synonymy and 153, 157, matUP 242 158, 268, syntactic use of 268, -tehl-eh and meaning 2, 4, 27, 39^7, 62, 91, 93, 96, 101, 163, 105, 158,169-74, 278 167, 236, 253, 266, 267, 269, 277, 278; adhikdra-wovds and, beliefs and 4, 62, 277, Gocarl 78-81 determined by the community 5, 56, guru 93-5 indeterminacy of 2-4, intersubjectivity and 27, 72, 93, naming and 40, of sentences 277, Harisvamin, date of 15-16 patterns of 1, 2, reference and 40, 47, single hermeneutics 6 sounds and 252, 253, 257, 263, 26,4, social hetunirdesa 99 anthropology and 2 n.2, theory of truth as holism 3 theory of 173 mehatiti 103-4, 168-9, 173; causal explanation ideal expressions 96, 168, 173-4, 273, 277 104,168-9 identification 179, 180, 181, 278 method 9-11 identity 180 interpretation 5, 7; Indian tradition and 7, Mlmamsakas 7, 274 methods of 7-8, models underlying 8 in satah 158-68; adhikdra-v/ords and 158, 162, Nagesa 45, passim interpretation of analyses in 162, 167, 276-7, naimittiki 69, 70 restriction on Mra&tf-interpretation 159-60, Nairuktas 27, 273; allegorical interpretations 167, ritual context of 166, sanctions definite and 27, 273, historicity and 27, 273 descriptions as true 167, 168, 173, selfNaigamakdnda 29 reference and 161, semantic nature of 168 Naighantukakdnda 28 ndmakarana 11*2-18, 169 ity apy asya bhavati 152-8, 171, 174, 268

Daivatakanda 29 Davidson, Donald 3 n.6, 5 definite descriptions 26, 161, 164, 166, 167, 173 Dikcarl 78-81 diksd 93-5 discourse 5-6, 95, 277; consistency of 6, myth as part of 7, nirvacana analysis and 277, universe of 5, 6 Durga 14, 23, passim; date of 14

General index

301

names 72; attribution of properties through 96, Para 59 descriptive 48, of Tantric deities 57, theory of Paramartha 34 n.48 meaning of 161 Parapara 59 nearness, see sdmlpya pdribhdsiki 69, 70 Nighantu 28-9, 31, 173; substitution principle paroksavrtti 35, 37 and'i73 Patanjali 10,13, passim] and the Nirukta 14 Nllakantha22-3 power 6, 86, 277; to determine correct usage nirdisyamdna-panbhasa 218, 219-29, 234-5, 187 268; necessity for non-grammarians 235, Powers (sakti) 58, 59, 60, 63, 78, 79, 80, 81, 89; partitive genitives recurring by anuvrtti and female 60, 78 226, 227, 228, relation to A 1.1.49 221, 231, prakdsa 79 234, relation to A 1.1.52, 54 and 55 211, 223, prasanga 176 n.4, 194, 217, 235, 238, 249, 255, 225, 228, 230-1, 235, tadantavidhi and 257, 262, 263, 264, 265, 266, 267 220-4, 227, 228, 231, yaddgama-paribhdsd prasiddhd 69, 70 and 220-4, 227, 228 Pratisakhya texts 175, 274 nirukta 27, 31; purpose of 29-30, relationship to pratyaksavrtti 36 vydkarana 32 prayojana 198-200 Nirukta 13, 27, 27-34; commentators on 14-23, properties: attribution of 96, 277, necessary or editio pnnceps of 24, Indo-European contingent 96, 167, relational 246 philology and 24-5, semantic analysis and 27, study of 23-5 Quine, Willard Van Orman 2, 3,4 n.8, 26; Niruktabhdsyatikd, date of 14-18 reform of language 26 Niruktaslokavdrttika 19, 20, 22-3 rasmir yamandt 101-2 Niruktavdrttika 18-22 reference 3, 72, 86, 96, 161, 162, 167, 277; nirvacana, definition of 27, 35 nirvacana analysis 8, 26, 27, 268, 269, 276, 277, inscrutability of 3, meaning and 40,47, power to determine 6, 277, truth and 3 n.6 278-9; alternative analyses valid 269-70, background of 25-7, beliefs and 277, relation 237, 238, 240-1, 246, 247; dnantarya epistemological device 277, establishes truth192, 213, 214, 217, 235, 241, 244, 'being in the value of sentence 277, general principles of place of 97, 162, 189, 'belonging to' 105, 35-9, grammarians' use of 32, in between linguistic elements 167, 173, 192, Ahirbudhnyasamhitd 57, in exegesis and 213, 234, between name and action or event metaphysics 57, model underlying it 8, 9, 98, between substitute and ideal expression 268-9, 276, purpose served by 33, 97, between term and explanatory expression relationship between various types of 168-74, 97, 98, 104, 159, 164, 168, 189, 236, 268, 269, semantic analysis 48, substitutional between terms in substitution operation 189, interpretation of 174, 268-9, theories of truth 190, between word and meaning 166 n.195, and 277, 278, through syllables and single 265, defined by sthdne 201, 213, 215, 221, sounds 89-93 238, m susbtitution model 236, marked by a nirvacanasdstra 13, 27, 279; Chinese Buddhist sthdnasasthi736-48, 268, rule A 2.3.50 and 237, sdmTpya 192, 213, 214, 217, 235, 241, 244 texts and 34-5, complement to grammar 32, considered a veddnga 27-8, continuous restriction, see niyama tradition of 23, imprints of in Kashmir 56, use Rjvartha, date of 14 in religious discourse 56, vydkarana and 25, 32, 75, Yaska's exposition of 35-9 sabdabrahman 254 Nisdtanatantra 68 n.35 sdbhydsa 124 nitya 165, 186, 270, 278 Sadasiva 58 n. 12 nityatva 259, 260, 264, 271, 272, 273 Saddanlti 32 nivrtti 249, 257, 259, 260, 262, 264, 265 Saiva Siddhanta 58-9,78 myama 197, 199, 201, 202, 203, 207, 211, 212 Saivism 57; Kapalika 58, Kashmirian 57, Kaula non-dualism 57, 58, 59, 60-1 59, 78, Krama 58, Tantric 57, Trika 59 Sakti, see Powers optionality 188 n.29 sdmTpya 192, 213, 214, 217, 235, 241, 244 Sanskrit 185; linguistic change and 187, model Pancaratra 57 speakers of 187 Panini 10, 13, passim sasthyarthdh, Nagesa's interpretation of pada 274-5; Mantrasastnc use of 274, 197-200' Vasubandhu's definition of 275 sdstra, Chinese translation of 34-5

302

General index

Sastra, methodological division in 183 tadantavidhi 220, 224, 225, 227, 228, 231, sat-159, 161, 163,166,167,173, 276, 277 nirdisyamdna-panbhdsd and 220-4, 227, 228, satT 187 231 Sayana 18 n 14, 29, 30, 31, 32, 33, 47 tantra 57 scripture 7, 57 Tantnc deities, names of 57 self-reference 250, 253, 266 tdntnka 57 Tarski, Alfred 3 n 6 semantics 3 n 6, 25, 97, 101, 265 sentences 50, 276, elliptical 276, meaning of tattva 25 277, semantic relations between 277, truth -teh and eh 104-31, 169, ablative attributes to conditions of 3, 50, truth-value of 277 123-31, ablative interpretation of 169-74, 278, genitive attributes to 112-23, genitive single sounds meaning of 252, 253, 257, 263, interpretation of 169-74, 278, relation to other 264, substitution of 252, 253, 266, 267 types of analysis 168-74 Sistas 187 temporality 272, 273 Sivatanusdstra 70, 88 Sivopadhyaya 89 text passages, presentation and citation of 11-12 Skanda-Mahesvara 14-18, 23, passim texts 1, exegesis of 1 Skandasvamin, see Skanda-Mahesvara time, notion of 186, 273 sphota 255 tirtha 55, 56 n 2 tradition 57, 84, 88, 95, 278, authorities of 95 sthana 176, 194, 196, 197, 201, 205, 207, 236, 238, 241, 248, 269, 271, 277, Bhartrhari on translation 2, indeterminacy of 4 250-5, in ritual context 278, Kaiyata on Tnka Saivism 59, 60, 78, incorporation of 255-8, kdla 'time' and 252, 253, 254, Krama elements in 78, ritual and 60 meaning of 217, 235, 238, 247, 248, 248-67, truth 3, 50, 95, 97, 173, 277, 278, mtersubjective 269, 277, 278, metaphorically considered a 5, mrvacana analysis and 277, presumption of relation 247, 248, Nagesa on 258-64, 95, theory of truth as theory of meaning 3 necessary condition for relation 194, 236, 241, n 6, 173, theories of 3 n 6, 277 244, 247, 248, 269, relational term 236, 244, truth conditions 97 247, 248 sthdnasasthT 153, 157, 158, 162, 173, 189-267, upacdra 248 n 169 268, 269, 277, identified by a mark 208-11, Upamsads 180, 181, 278 ordinary Sanskrit usage and 234, relation Usas 26, 171 marked by 207, 235, 236-48, 268, synonymy utsarga 185 and 153, 157, 158 vdcaka 79 sthdneyogd, interpretations of 192-7, 204-8 sthdmn 176-8, 225, 226, 233, 238, 269, 270, vdcya 79 271, 272, 273, 274, Albrecht Wezler on 176-8, vaidika 57 identified by nirdisyamdna-panbhdsd 225, Vaidyanatha Payagunda 216 226 varnasdrupya 92 sthdnivadbhdva principle 188-9, 226, 269, 270, vdrttika 21 273, 274, 275, in the ritual Sutras 188-9 vdrttikakdra 18-22, Saunaka as 21 stotd stavandt 102 Vasudeva Dlksita 214 substitutes 178, 187, whole words 187, 264, 266, Veda 7, 274, inaccessible parts of 69, 187, 274 meaning of 33, 46, understanding of 28, 31 substitution 158, 178, 179, 180, change and 278, vimarsa 73, 79 in Paniman grammar 175, m ritual Sutras 175, visayasaptamT 256 188-9, of single sounds 252, 253, 266, 267 voidness 81 substitutional model 96, 97, 173, 174, 175, 181, vydkarana 10, 13, passim 186-7, 188, 264, 269, 270, 277, 278, nonvyaktipaksa 222 n 112 temporality and 272, 273 vyatireka 200 Vyomavamesvarl 80 Svacchandabhairava 58 Vyomavyapimantra 275 Svacchandatantra 60-1, Ksemaraja's exegesis of 60-1, non-dualism and 60-1 svanta vowel 210 yaddgama-panbhdsd 220, 220, 221, 224, 225, synonyms 153,157, 158, 171, 173 227, 228, nirdisyamdna-panbhdsd and 220-4, synonymy 153, 157, 158, genitive case and 157, 227, 228 158, substitution criteria and 153 Yaska 13-14, passim, concerned with semantics syntactic units 275 31, 101, date of 13-14

University of Cambridge Oriental publications published for the Faculty of Oriental Studies 1 Averroes' commentary on Plato's Republic, edited and translated by E I J Rosenthal 2 FitzGerald's 'Salomon and Absal\ edited by A J Arberry 3 Ihara Saikaku the Japanese family storehouse, translated and edited by G W Sargent 4 The Avestan Hymn to Mithra, edited and translated by Ilya Gershevitch 5 The Fusul al-Madam of al-FdrdbT, edited by D M Dunlop (out of print) 6 Dun Karm, poet of Malta, texts chosen and translated by A J Arberry, introduction, notes and glossary by P Grech 7 The political writings of Ogyu Sorai, by J R McEwan 8 Financial administration under the Tang dynasty, by D C Twitchett 9 Neolithic cattle-keepers of south India a study of the Deccan Ashmounds, by F R Allchin 10 The Japanese enlightenment a study of the writings of Fukuzawa Yukichi, by Carmen Blacker 11 Records of Han administration Vol I Historical assessment, by M Loewe 12 Records of Han administration Vol II Documents, by M Loewe 13 The language of Indrajit of Orcha a study of early Braj Bhdsd prose, by R S McGregor 14 Japan s first general election, 1890, by K H P Mason 15 A collection of tales from Uji a study and translation of 'Uji Shui Monogatari', by D E Mills 16 Studia semitica Vol I Jewish themes, by E I J Rosenthal 17 Studia semitica Vol II Islamic themes, by E I J Rosenthal 18 A Nestonan collection of Chnstological texts Vol I Synac text, by Luise Abramowski and Alan E Goodman 19 A Nestonan collection of Chnstological texts Vol II Introduction, translation indexes, by Luise Abramowski and Alan E Goodman 20 The Synac version of the Pseudo-Nonnos mythological scholia, by Sebastian Brock 21 Water rights and irrigation practices in Lajh, by A M A Maktan 22 The commentary of Rabbi David Kimhi on Psalms cxx-cl, edited and translated by Joshua Baker and Ernest W Nicholson 23 Jaldl al-din al-Suyuti Vol I Biography and background, by E M Sartam 24 Jaldl al-din al-Suyuti Vol II'Al-Tahadduth binimat alldh\ Arabic text, by E M Sartam 25 Ongen and the Jews studies in Jewish-Christian relations in third-century Palestine, by N R M de Lange