Skip to content

Are Pāṇini’s sūtras descriptive or prescriptive sentences?

– Sergiu Al-George –

On the question of how Pāṇini’s sūtras are to be interpreted – as descriptive or prescriptive sentences – there has been considerable disagreement. A solution to this controversial matter is particularly important for understanding the logical structure of Pāṇini’s system.

The prescriptive status of Pāṇini’s sūtras has been supported by P. Thieme and L. Renou. The former considers the sūtras to be “grammatical injunctions” and observes that Kātyāyana and Patañjali “look upon Pāṇini’s grammar as a code containing the laws of correct word formation”. Accordingly, their “discussions have an analogon, not indeed in philological commentaries, but in a certain type of commentaries written by lawyers and law professors on codes of law”[1]. As for L. Renou, he qualifies the rules of the Aṣṭādhyāyī as “[tra] normatifs”[2]. This is the opinion of two scholars whose competence is strictly philological, untouched by any Western linguistic or logical prejudice against a normative grammar.

The opposite opinion was maintained for the first time by J.F. Staal in a well-known and frequently quoted article[3]. In contradistinction to P. Thieme and L. Renou, who did not advance any argument in support of their thesis, the American scholar argued extensively in favour of a descriptive status of Pāṇini’s sūtras. It should be noted that he did not acknowledge the opposite view. J.F. Staal’s major concern was to establish a clear-cut distinction between grammatical and ritual rules – as descriptive and prescriptive sentences respectively – with the purpose of demonstrating that the law of contradiction does not hold for both cases, but for grammatical rules only.

The American scholar approaches Indian thought within the framework of Aristotelian logic: according to the Stagirite, logic was concerned exclusively with descriptive-indicative sentences, because it is only such sentences that can be true or untrue (with respect to the subject/predicate relationship). This duality has been taken over by two-valued logical calculus, where a descriptive-indicative sentence is represented as F(x). J.F. Staal considers this an adequate formal expression for the interpretation of a Pāṇinian rule[4]. As for the treatment of ritual injunctions, he resorts to Aristotelian modal logic, wherefrom he retains only the “necessity” modality in order to render the imperative.

Anyhow, the Indian rules – be they ritual or grammatical – fall outside the narrow scope of a logic confined to descriptive-indicative sentences and truth values. In the past few decades – even before J.F. Staal’s article – Western philosophers have tried to systematize reasoning with propositions that have a temporal, interrogative or normative content. We are naturally interested in this last category, in the new logical theory of commands, also known as deontic logic. It deals with expressions such as: “ought (must)”, “may” and “must not”, which lead to the basic notions of “command”, “permission” and “prohibition”. Although this new branch of logic is still developing, it can even at this stage provide adequate criteria for a better approach and understanding of our subject-matter.

The new logical analysis of normative statements distinguishes between three major types of norm formulations. According to G.H. von Wright[5] there are:

  1. “rules”,
  2. “prescriptions” or “regulations” and
  3. “directives” or “technical norms”.

In turn, “prescriptions” fall under three categories:

  1. “commands”,
  2. “permissions” and
  3. “prohibitions”.

To the first type belong the rules of a game, grammatical rules and perhaps the rules of logic. Prescriptions are given or issued by a norm-giver (norm-authority) and are addressed to a norm-subject. As for permissive prescriptions, they have an intermediate status between commands and prohibitions. Directives or technical norms are concerned with the means to be used for the purpose of attaining a certain end; their standard formulation is a conditional sentence, whose antecedent specifies the end and whose consequent specifies the means that must or must not be used.

Let us now examine the arguments advanced by J.F. Staal in support of the descriptive status of Pāṇinian sūtras as against ritual rules (vidhi). He claims that Indian grammarians “do not use vidhi to denote positive linguistic propositions, but generally employ the term sūtra «rule» and sometimes, more specifically, utsarga «general rule»”[6]. This statement will appear strange to those conversant with the Mahābhāṣya, the most representative exegetical text. As for Pāṇini’s own text, vidhi is the only term employed to designate his own rules (I, 1, 56-58; I, 4, 13; II, 1, 1; VIII, 2, 2); he mentions neither sūtra nor utsarga. As for the inclusion of this term among the other values ascribed to li7, “optative suffix” (III, 3, 161), and by reccurrence (anuvr̥tti) to lot, “imperative suffix” (III, 3, 162), S.M. Katre quotes all these rules under the same entry, thereby not making any contextual distinction and invariably translating “a rule, formula, grammatical operation”[7].

A basic argument for assigning a descriptive status to Pāṇini’s sūtras is provided, according to J.F. Staal, by their syntactic form, different from that of ritual injunctions: whereas the latter contain a verb in the optative mood, the former are verbless: “As in India the science of grammar is primarily descriptive and not prescriptive, the grammatical sūtras are propositions where the verb occurs in the indicative mood or the construction is purely nominal”[8]. The same opinion is shared by H. Scharfe who believes that the purely nominal form of Pāṇini’s metalanguage invites a comparison with the Sanskrit noun phrase where a copula is implicitly understood[9]. Like G. Cardona[10], H. Scharfe refers to varttika 11 on II, 3, 1 and varttika 4 on II, 3, 46; but there the verb AS-, “to be”, ought not to be taken as a copula, having an existential value (not merely “is”, but “there is”). Disregarding this semantic difference, H. Scharfe concludes that: “most of them [i. e. sūtras] consist of only one clause with elementary constituents, which generally follow in this order:

  1. the  predicate of the term or feature under definition (description);
  2. the term or feature under definition (description);
  3. a conditioning factor;
  4. a connective”[11].

Although he assigns a descriptive status to Pāṇinian rules, the German scholar sometimes refers to them in normative terms, as for instance, “injunction”[12], “enjoining or prohibiting”[13]. Besides, in a previous work, where he considers the analogies between ritual and grammatical rules, as established by Patañjali, he agrees with O.A. Danielsson that niyama and vidhi “aus der Tradition der Rechts- und Sittenliteratur oder der Mimamsa stammen” [14].

The descriptive form of a sentence with the verb in the indicative is not incompatible with an actual normative meaning, as H.G. von Wright observes:

It must not be thought that imperative and deontic sentences are the only grammatical types of sentence which are used as norm-formulations. Indicative sentences, other than deontic sentences, are also quite commonly used for expressing norms[15].

With regard to the Sanskrit texts, the same fact was pointed out by L. Renou:

dans le Rituel comme dans le Coutumier description signifie norme, ce qui est est ce qui doit être; on peut donc indifféremment poser l’un ou l’autre modes[16].

In other words, normativeness is not strictly related to the form of the sentence, but to its context and purport, as Bhartr̥hari remarked a long time ago:

The purport of the expression (vācyatā) does not lie in the form of the employed expression[17].

This is the reason for the fact that Staal found paradoxical, namely that Śaṅkara exemplified injunction by sentences where the verb occurs in the indicative mood[18].

The verbless form of Pāṇini’s sūtras presupposes both the elision of the verb and the setting up of peculiar syntactic values for declensional suffixes, specific to his metalanguage. Under the abbreviated form of most rules, there usually lies a complex sentence; such a complex sentence expresses – as we shall see below – a relation from antecedent to consequent. Leaving codification rules aside, the others fall into two categories: the first category includes all the rules that establish a relation between two linguistic entities belonging to different levels, such as that between a semantic entity and a morphological one; the second category is that of substitution rules. In the first category the antecedent (protasis) is expressed by means of a noun in the locative case, which Patañjali calls viṣayasaptami, “field locative” or “the locative of range”. What we have here is a locative absolute, where sati – the present participle of the verb “to be” (AS-) – is elided. O. Böhtlingk is the first to have taken the viṣayasaptami as a locative absolute[19]. Consequently, the simple locative may play the syntactic part of a locative absolute expressing as such conditional or concessive clauses. Later Pāṇinian exegesis confirms this interpretation of the viṣayasaptami: Haradatta was the first to call this locative satsaptami, “the locative of «where there is»”. At the same time, Haradatta showed that the satsaptami is stipulated in Aṣṭādhyāyī II, 3, 37[20].

Thus, all the rules resorting to the viṣayasaptami, where the verb “to be” is to be supplied, are in fact complex sentences where the locative is equivalent to a protasis. In the above‑mentioned rule (II, 3, 37) – without denominating in any way the locative absolute – Pāṇini ascribes it the value of lakṣaṇa, “indication”. As such, apodosis expresses the indicated operation, and the whole sūtra should be understood under the general form: “when there is A, then B is indicated”[21]. We have to point out that the viṣayasaptami is present not only in grammatical rules, but also in metagrammatical rules, paribhāṣā, axioms where an operation of interpreting and applying a grammatical rule is indicated. The frequent occurrence of the viṣayasaptami in the oldest axiomatic syntax of the paribhāṣās is worthy of notice: over a third of their list – as given by Nāgojībhaṭṭa – resorts to this syntactic form. Unlike the viṣayasaptami, which does not exclusively belong to Pāṇini’s syntax – for it can be found in other technical texts – another value of the locative, which is peculiar to Pāṇini’s metalanguage, occurs in the Aṣṭādhyāyī. In rule I, 1, 66, Pāṇini stipulates that in a substitution rule the word in the locative is placed after the substituendum. Patañjali calls it parasaptami, “the locative of the following one”. This function of the locative case is symmetrical to that laid down for the ablative case in sūtra I, 1, 67: the word in the ablative is placed before the substituendum. In substitution rules, both the locative and the ablative, representing circumstantial criteria for the substitution operation, play the part of an indicational protasis, in a way similar to the viṣayasaptami; still, instead of “when there is A”, we shall have respectively “when A follows X” and “when A precedes X”. As for the apodosis – which mentions the operation of substitution – the fact that this operation is indicated is shown precisely by the Sanskrit term denoting the substitute: adeśa means literally “the indicated one”[22]. We may therefore conclude and, at the same time, emphasize that – with the exception of the saṁjñāsūtras, where a denotational definition is given – the general syntactic structure of the sūtras cannot be assimilated to a noun-phrase with one subject and one predicate; it really is the sentence with a protasis and an apodosis. This syntactic dichotomy substantiates itself in the distinction made by post-Pāṇinians between prasaṅga, “the field of virtual application”, and prasakta, “that which is virtually applicable”[23]; these correspond respectively to the protasis and the apodosis. The semantics of both terms shows that the rules are not conceived in the existential mode (true/false), but in the virtual one, which expresses only a provisional application or validity, liable to be cancelled by a subsequent rule or by factual conditions. Their virtuality is due to their prescribing the execution of a technical operation. In such cases the norm is not categorical, but subject to a condition[24]. Characterized as a rule, the norm itself is expressed in the apodosis which shows the technical operation, what is to be done (kārya); it is performed only when the entity undergoing the operation (ryin) is no longer virtual but actual[25].

The operation as subject to a condition is encountered in ritual rules (vidhi), which are not mere injunctions but injunctional rules. Ritual rules, therefore, are also technical norms: they express a means of attaining a certain end, and the normativeness of the means is clearly indicated by the optative or imperative mood.

The grammatical rule differs from the ritual one by its more technical form, where the verb and its mood are missing. But in both types of rules the antecedent represents the criterion of validity of a certain operation which must be performed (kārya), be it sacrificial or grammatical. In ritualism, Prabhākara designates the normative element in itself – as different from its conditional form (vidhi) – by niyoga[26]; in grammar this corresponds to karmasādhanā as distinguished from bhāvasādhanā (Patañjali on I, 1, 58, init.). These two terms indicate the ambiguity of the word vidhi in the grammatical vocabulary: it can either mean

  1. karmasādhanā, glossed yo vidhiyate, “that which is prescribed”, or
  2. bhāvasādhanā, “the process of prescribing”, glossed vidhana.

The exegesis lying behind these two words is best summarized by Y. Ojihara and L. Renou[27]. The profound insight behind these refinements becomes obvious when we consider the treatment of negative markers in grammatical rules.

In a descriptive-indicative sentence the negative markers operate within the duality assertion/negation, in direct relation to the bivalence true/false. In a normative sentence, negative markers operate in a dissimilar and more complex way: instead of the duality of descriptive sentences, we are dealing here with command, prohibition and permission, to which truth-values cannot be assigned. Negative markers play a prominent part in the Aṣṭādhyāyī, where they are abundantly resorted to; as such, their treatment, as well as their employment – either as verb negation or noun negation –, provide a considerable amount of data relevant to semantics, logic and semiotics[28]. The noun negation is called by commentators paryudasapratiṣedha, as distinguished from prasajyapratiṣedha, verb negation. We shall dwell upon the latter, which alone lend itself to a new treatment in the light of normative logic.

As Patañjali observes (on I, 1, 5, after varttika 3), unlike noun negation (privative a-, “un‑”), which occurs both in the protasis and the apodosis, verb negation (na, “not”) operates only in the apodosis, where it suppresses the prasakta, “that which is virtually applicable”. L. Renou renders prasajyapratiṣedha by “prohibition valable apres application (virtuelle)”[29], thus acknowledging its normative status. On this topic, G. Cardona’s treatment seems rather confused. On the one hand, it seems that implicitly he adopts a prescriptive treatment: for instance, when he accepts Patañjali’s definition of vyākaraṇa as “word instruction” (śabdānuśasana), when he admits that the rules provide operations[30], or when he assserts that “the general rule prescribes”[31]. On the other hand, he seems to adopt a descriptive treatment when he renders prasajyapratiṣedha by “negation” instead of “prohibition”. This translation is implied in his own postulation of the verb “to be” (“there is”) in the verbless form of the sūtras[32]. Accordingly, the meaning of na is “absence”[33], namely “the absence of the operation”[34]; but such an absence cannot be – at the present stage of logic – the result of an existential negation, which is descriptive; for an operation can be suppressed only by a prohibition.

As we have already seen, it is only in the sphere of prescriptive sentences, between assertivity and negativity, that a third possibility can arise, namely the permission to opt. This type of permission occurs in Pāṇini’s sūtra I, 1, 44: na veti vibhāṣā, “«May or may not» is called option (vibhāṣā)”. A rule containing the expression na va is to be interpreted as expressing permission to choose between two operations prescribed under the same condition; as such, it represents what deontic logic calls “disjunctive permission”[35]. The status of the grammatical vibhāṣā is identical with that of the ritual vikalpa, an identity provided by the synonimy vibhāṣā = vikalpa, which is sometimes encountered in the Mahābhāṣya, and more often in the Kaśikavr̥tti. As for the impossibility to assign a descriptive status to Pāṇini’s optional rules, we should bear in mind that Śaṅkara[36] – quoted by J.F. Staal[37] – insisted that option (vikalpa) cannot pertain to descriptive-indicative statements:

one does not exercise an option as to whether a thing is thus or not thus, or as to whether it exists or not.

The independent status of permissive sentences as against enjoinments or prohibitions has been the object of many debates and has led to similar interpretations in both ancient Indian grammatical texts and contemporary normative logic: Kātyāyana’s assertion, that in a vibhāṣā a prohibition is superseded by an option (on I, 1, 44, varttika 3), anticipates by two millenia G.H. von Wright’s interpretation of permission “as nothing but the absence or non-existence of «corresponding» prohibitions”[38]. Contemporary normative logic might discover in ancient Indian texts – both ritual and grammatical – a substantial and rigorous treatment of permissive prescriptions.

Finally, let us reconsider the law of contradiction as it has been treated by J.F. Staal in ritual and grammatical rules. He focuses on the word vipratiṣedha, which he translates as “mutual prohibition” or “contradiction”[39]. This re-examination aims at establishing whether vipratiṣedha should be translated by “contradiction” or by “mutual prohibition”, given that these two concepts pertain to different logics; there is need to distinguish clearly between the two different meanings and to choose one to the exclusion of the other. After considering the word vipratiṣedha in ritual texts, where it “holds in cases of mutual contradiction between two rules applicable to the same situation”[40], J.F. Staal reaches the following conclusion:

While the law of contradiction need not necessarily hold for injunctions which are always injunctions enjoining action, the same law does hold for the results of the activities based upon these injunctions[41].

When he considers the same word (vipratiṣedha) in the Aṣṭādhyāyī, where it occurs in the famous paribhāṣāsūtra I, 4, 2:

Where there is mutual prohibition [between two different rules], the latter is to be applied[42],

his treatment is different. Using mathematical formalization, J.F. Staal concludes that the formulation of the respective paribhāṣā “presupposes the validity of the principle of contradiction”[43] or, as he further explains, “in grammar the law of contradiction is immediately applied to propositions”[44].

It is not our intention to re-examine the logical calculus used to prove that whereas in the case of grammatical rules the law of contradiction holds at the rule level, in the case of ritual ones it concerns only the prescribed operation. A careful and complete examination of the Mahābhāṣya renders such a re-examination useless for the purpose of questioning J.F. Staal’s conclusions. In commenting on sūtra I, 4, 2, Patañjali starts by defining vipratiṣedha as a mutual prohibition[45]; Kātyāyana makes this situation explicit:

Where there are two virtual applications (rules) with different meanings at the same place, there is vipratiṣedha[46].

Patañjali, in his turn, adds: “At the same place and at the same time”[47], and illustrates this situation by rules VII, 3, 102 and 103; the prevalence of the latter over the former leads to the correct form vrksebhyah instead of vrksabhyah. Staal closely follows the Mahābhāṣya‘s analysis up to this point[48], but the very meaning of the paribhāṣā is revealed further down in the commentary. Thus, Kātyāyana states that the application at the same place of two different rules is an “impossibility” (asambhāva), but that their virtual application is not affected[49]. Patañjali indicates that the impossibility in question – at the same place and at  the same time – concerns the two stipulated operations (kārya), whereas prasaṅga, “virtual application”, concerns the two sūtras[50]. His analysis, where vipratiṣedha, as impossibility, concerns only the operations prescribed by the rules, and not the rules themselves as virtual statements, anticipates in fact – although in more precise terms – J.F. Staal’s own analysis of a possible “contradiction”, not between ritual injunctions but only between “the activities based upon these injunctions”. Of course, in both ritual and grammatical rules, vipratiṣedha has nothing to do with contradiction. Contradiction is the incompatibility between two descriptive statements, obvious in the very structure of their expression, which differs according to the presence or absence of the negative marker. Where normative formulations are concerned, contradiction is ruled out: here the incompatibility cannot concern the rules themselves, but only their actual application. Normative logic calls this kind of incompatibility “imperative incompatibility” or “prohibition to carry out at the same time p and q”; it is represented as: ! ù (p Ù q). This is recognizably Patañjali’s own formulation, to which he would have only added “in the same place”.

Admittedly, Pāṇini’s grammar offers a complete description of the Sanskrit language, but he only describes as much as he prescribes. The underlying regularities of Sanskrit are formulated as a set of rules, but these are identical to the ones pertaining to its own technical devices of codification. In Western linguistics the concept of normativeness has been compromised, as unscientific, by the old normative grammars, and our logic has not concerned itself with it until recent decades. However, the significance of normativeness in linguistics has been indirectly reappraised by information theory and semiotics. These new humanities disciplines have restored the concept of code and developed it to such an extent that it is now shared by  the methodology and philosophy of biological sciences. We may therefore say that normativeness in Pāṇini’s system can no longer be considered unscientific; on the contrary, Pāṇini is relevant to contemporary linguistics and philosophy, his work being the first semiotic codification conceived with unparalleled rigor and complexity.

It seems that Pāṇini continues to be for Western culture a constant anticipation, that the permanent rediscovery of his work will never end as long as Western culture will continue to develop new branches of study in the field of language.


 [1]. P. Thieme, “Pāṇini and the Pāṇiniyas”, JAOS 76 (1956), 1, pp. 22, 23.

 [2]. L. Renou, “Sur le genre du sūtra dans la littérature sanskrite”, JA 251 (1963), 2, p. 169.

 [3]. J.F. Staal, “Negation and the law of contradiction in Indian thought: a comparative study”, BSOAS 25 (1962), 1, pp. 52-71.

 [4]. J.F. Staal, op. cit.,  p. 60.

 [5]. G.H. von Wright, Norm and action. A logical enquiry, London 1963, pp. 6-15.

 [6]. J.F. Staal, “Negation and the law of contradiction”, p. 64.

 [7]. S.M. Katre, Dictionary of Pāṇini II, Poona 1968, p. 517 (s.v. vidhi).

 [8]. J.F. Staal, op. cit., p. 56.

 [9]. H. Scharfe, Pāṇini‘s metalanguage, Philadelphia 1971, p. 17.

 [10]. G. Cardona, “Negations in Pāṇinian rules”, Language 43 (1967), 1, p. 39.

 [11]. H. Scharfe, op. cit., p. 39.

 [12]. Ibid., p. 46.

 [13]. H. Scharfe, Pāṇini‘s metalanguage, p. 32.

 [14]. H. Scharfe, Die Logik im Mahābhāṣya, Berlin 1961, p. 38, n. 2.

 [15]. G.H. von Wright, Norm and action, p. 101.

 [16]. L. Renou, “Sur le genre du sūtra”, p. 183.

 [17]. VP III, 3 (sambandhao), 26: na ca vacakarūpena pravr̥ttasyasti vācyatā (apud M. Biardeau, Théorie de la connaissance et philosophie de la parole dans le brahmanisme classique, Paris 1964, p. 422).

  [18]. J.F. Staal, “Negation and the law of contradiction”, p. 62.

  [19]. O. Böhtlingk, Pânini’s acht Bücher grammatischer Regeln II, Bonn 1840, p. 66.

  [20]. Y. Ojihara and L. Renou, La Kaśika-vr̥tti III, Paris 1967, p. 47, n. 2.

  [21]. See supra, p. 111.

  [22]. MBh ad I, 1, 62, after varttika 14: adiśyate yah sa adeśah /

  [23]. L. Renou, Terminologie grammaticale du sanskrit, Paris 21957, pp. 229-230 (s.v. prasakta).

  [24]. G.H. von Wright, Norm and action, p. 171.

  [25]. MBh ad I, 1, 1, after varttika 7: sato hi karyinah karyena bhavitavyam (p. 40, l. 4). See L. Renou, Terminologie, p. 130.

  [26]. M. Biardeau, La philosophie de Mandana Miśra vue à partir de la Brahmasiddhi, Paris 1969, p. 62.

  [27]. Y. Ojihara and L. Renou, La Kaśika-vr̥tti III, p. 88, n. 3.

  [28]. See S. Al-George, Limbă și gîndire în cultura indiană, București 1976, pp. 114-134.

  [29]. L. Renou, Terminologie, p. 230.

  [30]. G. Cardona, “Negations in Pāṇinian rules”, p. 35.

  [31]. Ibid., p. 39.

  [32]. Loc. cit.

  [33]. Ibid., p. 35, n. 2.

  [34]. Ibid., p. 41.

  [35]. G.H. von Wright, Norm and action, p. 160.

  [36].  Œaṅkara, Brahmasūtrabhaṣya I, 1, 2: na tu vastv evam naivam asti nāstīti vā vikalpyate (ed. N. R. Acharya, Bombay 31948, p. 8, ll. 13-14).

  [37]. J.F. Staal, “Negation and the law of contradiction”, p. 62.

  [38]. G.H. von Wright, Norm and action, p. 85.

  [39]. J.F. Staal, “Negation and the law of contradiction”, p. 53.

  [40]. Ibid., p. 53.

  [41]. Ibid., p. 61.

  [42]. Pan I, 4, 2: vipratiṣedhe param kāryam //

  [43]. J.F. Staal, “Negation and the law of contradiction”, p. 56.

  [44]. Ibid., p. 63.

  [45]. MBh ad I, 4, 2, init.: itaretarapratiṣedho vipratiṣedhah / anyo’nyapratiṣedho vipratiṣedhah //

  [46]. MBh ad I, 4, 2, varttika 1: dvau prasaṅgav anyarthav ekasmin sa vipratiṣedhah //

  [47]. MBh ad I, 4, 2, after varttika 1: ekasmimś ca yugapat.

  [48]. J.F. Staal, “Negation and the law of contradiction”, p. 55.

  [49]. MBh ad I, 4, 2, varttika 2: ekasmin yugapad asambhāvat pūrvaparaprapter ubhayaprasaṅgah //

  [50]. MBh ad I, 4, 2, after varttika 2: kāryayor yugapad asambhāvah śāstrayor ubhayaprasaṅgah //

Published inUncategorized