Does Asymmetric Signification Rely on Conventional Rules? Two Answers from Ancient Indian and Greek Sources *

The topic of asymmetry between the semantic and the phono-morphological levels of language emerges very early in Indian technical and speculative reflections as it also does in pre-socratic Greek thought. A well established relation between words and the objects they denote (the so-called one-to-one principle of correspondence) seems to have been presupposed for each analysis of the signification long before its earliest statement. The present paper aims at shedding light on two different patterns of tackling the mentioned problem. The first approach sees asymmetry as an exception to the regular correspondence between language and reality, whereas the second approach considers language in itself as a conceptualisation which does not faithfully represent reality. In the latter case, asymmetry is no longer an exception, but the rule. Traces of a ‘heterodox’ concept of Conference of Bravery. The Figure of the Hero in Indian Literature, Art and Thought, Cagliari ā tya th


Introduction: the Alleged Symmetry between Words and Objects
The present joint paper focuses on the topic of asymmetry between the semantic and the phonomorphological levels of language which emerges very early on in Indian technical and speculative reflections, as it also does in pre-socratic Greek thought. Our shared research aims at shedding light on two different patterns of explaining such a linguistic phenomenon. The first approach sees asymmetry as an exception to the regular correspondence between language and reality, whereas the second approach considers language itself as a conceptualization that does not faithfully represent reality, and hence, asymmetry is no longer an exception, but the rule. Before dealing with asymmetry, we need to take a short step back and depict a remote and common background where the symmetry between words and the objects they denote constitutes a given datum. In fact, these two opposed historical interpretations, in which linguistic asymmetry was either a natural or a conventional exception, at a certain point in time, actually derived from the reflections on this alleged symmetry between words and objects.
The first problematic way of considering asymmetry as an exception depends, in our opinion, on a presupposed basic symmetry of language, namely, on a sort of one-to-one principle of correspondence between words and the objects they denote 1 -which we assume was presupposed both in ancient India and in ancient Greece. This principle is clearly expressed in the Pāṇinian grammatical tradition only from the 3 rd century BCE onwards. 2 According to Kātyāyana, words as a rule apply per object: one and only one word-form matches with one and only one object. 3 To sum up, word-forms (śabda) definitely play the role of causes in the Pāṇinian framework and give rise to the cognition of objects (artha) in the mind of the participants in the communicative event. The addresser actually needs to employ words, for instance, to utter them, in order to arouse the relevant mental image in the mind of the addressee. Thus, the physical perception of words can really give rise to the relevant concepts.
Of course, we are keen to discover more about the previous speculative scenario in both India and Greece, which has brought us to preliminarily concentrate on some evidence that, in our opinion, proves that both Indians and Greeks were also convinced of this correspondence before the third century BCE.
The ancient popular etymologies included in Vedic sources-or better, the etymologizing stylistic figures entailing an undeniable magico-linguistic intention on the part of the poets, as they are presented by Deeg 14 -seem to rely on this belief. Already in the Ṛgveda [ṚV] 15 and in the Atharvaveda, whose earliest hymns may date back to the twelfth century BCE, the paretymological connection between theonyms and the specific role played by the matching Gods is rooted in this principle of denotative integrality.
For instance, the ancient Vedic Saṃhitās regularly explain the etymological connection between the noun which denotes fire-Agni-and the nominal base which means "the first, that which is in front." Analogously, pṛthivī (the earth/the Goddess Pṛthivī) is technically explained as "the wide [earth]" merely because of the link with the verb prath-"to spread" in Nir 1.14: prathanāt pṛthivīty āhuḥ, "They call it pṛthivī on the basis of the verb 'to spread.' Of course, a more interpretative translation could be: "They call it pṛthivī because it has been spread out," 21 since this name is the object of comparable etymological figures of speech in both the ṚV and in the AVŚ, where a specific agent, such as Indra, is often even singled out for the action (denoted by the verb prath-'to spread out') which is linked to the analysed noun/name (pṛthivī). 22 Nevertheless, Nirukta explicitly refuses excessive speculation on the motivation of the single relations between words. Instead, preference is given to a regular frame of correspondences among words, within which the single meanings can generally be smoothly detected.
Nir. 1.14: ka enām aprathayiṣat kim ādhāraś ceti. atha vai darśanena pṛthuḥ. aprathitā ced apy anyaiḥ. athāpy evaṃ sarva eva dṛṣṭapravādā upālabhyante "'But who spread it, and what was the base?' (We reply that) it is indeed broad to look at, even if it is not spread by others. Otherwise, in this way, all known words can actually be found fault with." It is therefore a perceptible fact that the earth is wide and this in itself must suffice. Little does it matter if we do not know the diachronic story of this linguistic usage or the aetiological myth. A markedly technical and scientific stance in highlighting this kind of linguistic connection is thus intentionally inaugurated. Somehow, mere perception (pratyakṣa), 23 warrants the enunciated relation between the object of language and its denotation-in this case between the quality of being wide and the substance earth that possesses it, in other words between the qualifying word and the qualified object. The Earth is wide by nature, and by nature it has to be called "the wide one" per antonomasia. It cannot just be a convention.
Thus, a scientific method slowly originates from a poetic and stylistic pre-scientific way of focusing on the principles of signification. Visible items are linked with audible items, because two audible items that denote two comparable objects are, in turn, also comparable. Thus, language seems to be a reliable means of knowledge and its reliability can be proven on the basis of perception. We shall see that this principle of correspondence is not explicitly expressed in archaic Greek sources, but rather, it seems to be presupposed, precisely because it is questioned by the authors. On the other hand, there is actually less distance between the aforementioned See, for instance, Lātyāyanaśrautasūtra 33 8.9.1-4: upahavye devatānāmadheyāni parokṣaṃ brūyuḥ svasthānāsu. pratyakṣam asvasthānāsu. devaśabdaṃ sarvatra varjayeyuḥ. hotā devo mahīmitrasyeti hotā yajñai mahīyajñasyeti brūyur iti somam, "During the upahavya, they should pronounce the names of the divinities cryptically in their respective places. In places other than their original one, they should pronounce them perceptibly. They should avoid the word deva in all circumstances. In place of the two words deva and mitra in the hotā devo mahīmitrasya text, they should say hotā yajñā mahīyajñasya; instead of soma they should say indu." Nevertheless, no doubts seem to arise regarding the reliability of both the (perceptible and cryptic) denotations as a valid means of knowledge, since the noun indu somehow depicts a real facet of soma, in the same way as Śakra is a truthful epithet for the god Indra. This particular example of asymmetry is not unmotivated at this step of the reflection on language.
Mylius considers the origin of this tradition as Sāmavedic, based on a sort of foundation myth of this rite, recounted in another Sāmavedic work (Jaiminīyabrāhmaṇa [JB] 2.150). 34 In fact, JB section 2.148-150 is entirely devoted to the explanation of the cryptic upahavya given by Prajāpati to the gods as a sort of weapon against their adversaries. The involvement of original epithets, synonyms, Kenningar, and other stylistic devices tuned to make the comprehension of poetry indirect has been recognized in the earliest sources of several Indo-European languages. Thus, the afore-mentioned Kalpasūtra and Brāhmaṇa explanations of this kind of cryptic usage of language might be the rationalizing (prescriptive or mythic) re-interpretation of the main aim of the ritual-poetic contest. 35 This might have also consisted in inventing new metaphorical or metonymic designations of the subjects in question, in order to defeat their adversaries and receive the deserved reward from the patron of competitions. 36 Therefore, such a type of exception to the one-to-one relation between word-forms and their denotations seems to be classified as an option available in the very nature of language itself, even though sometimes it is the mere ephemeral fruit of individual creation.
The most ancient actual discussion on asymmetry is proposed later and ends in favour of its natural encompassment. The relevant section (Nir 1. [12][13][14] is devoted to the specific derivational relation between an action and the matching nomen agentis. The starting point consists in maintaining that nouns are derived from verbs. Nir 1.12 nāmāny ākhyātajānīti śākaṭāyano nairuktasamayaś ca, "Nouns are derived from verbs according to Śākaṭāyana and according to the communis opinio of etymologists." of this rule to the mere lexical repertory of derived nouns governed by grammatical rules, such as go "cow," aśva "horse," puruṣa "person" and hastin "elephant." 37 There are several reasons to argue that Śākaṭāyana's thesis has to be refuted. Yāska lists six of Gārgya's objections, 38 among which three actually pertain to the problem discussed here: (Nir 1.12) atha cet sarvāṇy ākhyātajāni nāmāni syur yaḥ kaśca tat karma kuryāt sarvam tat sattvaṃ tathācakṣīran. yaḥ kaścādhvānam aśnuvītāśvaḥ sa vacanīyaḥ syāt. yat kiñcit tṛdyāt tṛṇaṃ tat, "If all nouns are derived from verbs, each individual who performs a certain action should be called by the same name. Whosoever runs on the road should be called 'runner' (aśva). Whatever pricks [should be called] 'pricker.'" (Nir 1.12) athāpi cet sarvāny akhyātajāni nāmāni syur, yāvadbhir bhāvaiḥ saṃprayujyeta, tāvadbhyo nāmadheyapratilambhaḥ pratilambhaḥ syāt. tatraivaṃ sthūṇā daraśayā vāṃ saṃjanī ca syāt, "Moreover, if all nouns are derived from verbs, [each individual] should obtain as many names as the actions with which it is connected. Thus a column should also be 'that which has been fixed in a hole' and 'that on which one hangs (sañj-) [something].'" (Nir 1.13) athāpi ya eṣām nyāyavān kārmanāmikaḥ saṃskāro yathā cāpi pratītārthāni syus tathaināny ācakṣīran. puruṣaṃ puriśaya ity ācakṣīran. aṣṭety aśvam. tardanam iti tṛṇam, "Furthermore, each grammatical form should be regularly named after the relevant [denoted] action, and these [objects] should be called in such a way that their denotation has to be clearly understood: puruṣa 'person' should be called puri-śaya, lit. 'he who lives in the city,' aśva 'horse' should be called aṣṭṛ 'he who covers (distances)' and tṛṇa 'grass' should be called tard-ana 'that which pricks.'" Yāska's answers are all inspired by the same principle: only linguistic usage has to be taken into account, and no attention should be paid to other sophistries. For instance, the first answer is as follows (Nir 1.14): Therefore, we can be sure that in Nirukta's age, asymmetry, especially between nomina agentis and the actions which they imply, was a well-known and accepted fact. It was probably considered as a natural part of language, which better emerged when its functioning was analyzed from the synchronic point of view. On the contrary, the actual correspondence between deverbal nouns and the paretymologically linked verbal bases, when it was strictly grammatically governed, 39 was considered quite obvious and uninteresting, so that the basically supposed oneto-one relation between word-forms and their objects was used as a purely ideal schema.
Therefore, according to the sixth-fifth century BCE testimony supplied by Nirukta, the link between word-forms and their objects can, by nature, also be an asymmetric relation, but it does not depend on artificial rules. In fact, since each object to be denoted is indeed multifeatured, it can consequently be denoted by more than one noun. Thus, the natural meaningform relations are strenuously defended. Three centuries later, Pāṇini's commentators seem to be consistent with this perspective. In Kātyāyana's opinion, the linguistic form has to be taken as a whole, rather than by inspecting its single parts. As a consequence each word is svābhāvika "autonomous" ("grounded on its intrinsic conditions"), rather than "conditioned by recognizable factors or causes" (be they external or internal).
Both vārttikas occur in discussions arising from some difficulty in reconstructing the denotation of a synthetic form by the analytical denotation of its constituents, thus in facing some specific exceptions to the aforementioned linguistic one-to-one relation. For instance, these vārttikas explain phenomena such as the occurrence of several objects of the same polysemic word akṣa which conveys the sense of "1. axle 2. rosary seed 3. die" or even the denotation of both mother and father, by means of the unique dual form of the noun commonly conveying the single sense "father" (pitarau), or the asymmetry between a root noun and a suffixal noun which share the same role of nomen agentiscf. baladā-"one who gives strength" vs. kanyā-dātṛ-"one who gives (a 'giver' of) a daughter in marriage." 41 The unpredictable signification mechanism is explicitly labelled as a natural fact, svabhāvika, autonomous from grammatical rules and ultimately from human efforts. This is similar to Yāska's conclusion when he says: language includes this form and this alone must suffice. More technically, Kātyāyana concludes that the denotation of an inflected word (pada) has to be defined as svābhāvika (independent from its morphs). 42 Moreover, Patañjali went deeper into the consequences of this detected linguistic mechanism, which, in his opinion, could indeed result in a risky association, if it were purely a conventional matching between objects and the word-forms which can denote them. In other words, the mutual comprehension between speaker and their interlocutor could not indeed be guaranteed. 43 Thus, Patañjali maintained that a mere grammatical device taken apart from its actual linguistic usage could not ensure a sufficient shared and understandable denotation. Likewise, a rule stating that a cow has to be called a horse and vice versa should be in manifest contradiction with ordinary usage which is consequently so important. Therefore, the symmetric schema of signification might have been evaluated as a sort of basic grammatical convention. This was used in order to easily arrange linguistic knowledge, although grammarians themselves were well aware that the patterns of signification were actually asymmetric by nature. The descriptive pattern of linguistic reality was thus assumed to be basically symmetric, so that a permanent (nitya) relation between the words of the Sacred Texts and their meanings could be taken as granted. Nonetheless, self-evident exceptions had to be admitted and merely registered (on the basis of common usage), though not explained by means of grammar. 44

In Ancient Greek Sources
Even if the cultural context as a whole is very different, the development of Greek thought about the word-object relation and about asymmetry-such as polysemy and synonymy-is surprisingly similar to that of Indian culture: both start by considering the word-object relation as biunique and well-established, 45 but it is subsequently conceived as a human undertaking and as a convention of which the asymmetry is part. As far as the Greek sources are concerned, the aim of the present contribution is to show how these two opposite conceptions developed in presocratic thought. In addition, we would like to demonstrate that they are bound to the problem of language as a means of knowledge, from the first occasional reflections on language in the early literature, 46 to the more systematic ones of the philosophers and sophists. 47 The archaic Greek thought on language conceives the word-object relation as responding to a sort of one-to-one principle of correspondence, which presupposes an ontological link between a name and denoted object: names, if well intended, are thus capable of revealing the very nature of things. In fact, archaic sources show that the oldest popular etymologies are the means by which the essence of an object can be revealed through the analysis of its name. Although paretymologies find origin in mythological thought, seeing that they were applied from an early age to the names of heroes and gods, they are also the first evidence of a pre-scientific way of thinking about language. The paretymology of the name of Ἀστυάναξ (Hom. Il. 6.478, Ἰλίου ἶφι ἀνάσσειν) or of Ἀφροδίτη (Hes. Th. 195-198, ἐν ἀφρῷ | θρέφθη) are striking evidence of this. 48 In the Homeric poems there are some occurrences of double denominations, which can be divided into three types: 49 1. Double denominations in general (patronymics, double names of animals, objects, deities The last example shows that etymology is used in the Homeric poems to explain one of the two proper names of a human character. 52 Hesiod uses it to exclusively explain the names of deities and never opposes the names given by gods with those given by humans. 53 In line with the aim of the Theogony, which tries to put the traditional divine universe in order, Hesiod associates only one name to each deity, as in the case of Briareus and Ino, who, instead, are characterized by two names in the Homeric poems. 54 The examples quoted above show that the Greeks had already begun to observe the existence of asymmetry in language in the archaic age. If the double denomination involves common nouns, this fact is merely recorded. However, if it involves proper names, it is perceived as particularly anomalous and in need of an explication. This is well exemplified in fr. 7 B 1 DK of Pherecydes of Syros, considered by the ancient sources as the author of the first Greek literary text written in prose: Ζὰς µὲν καὶ Χρόνος ἦσαν ἀεὶ καὶ Χθονίη• Χθονίῃ δὲ ὄνοµα ἐγένετο Γῆ, ἐπειδὴ αὐτῇ Ζὰς γῆν γέρας διδοῖ. "Zeus, Chronos and Chthonie were eternal; but the name of Chthonie became Ge, because Zeus gave her the earth as an honorific gift." Pherecydes' aim is to counteract the contradictory idea of the deities who are said to be "forever," but at the same time, "born" (as happens, for instance, in Hesiod) by demonstrating that it is only their name that changes. 55 He explains Chthonie's new name by means of a paretymology 56 which connects it with her being attributed the domain of the earth. We can also find this kind of justification in the "one deity-two different names-two religious domains" scheme, well exemplified by fr. 116 Kahn 57 of Heraclitus: Εἰ µὴ Διονύσῳ ποµπὴν ἐποιοῦντο καὶ ὕµνεον ᾆσµα αἰδοίοισιν, ἀναιδέστατα εἴργασται• ὡυτὸς δὲ Ἀίδης καὶ Διόνυσος ὅτεῳ µαίνονται καὶ ληναΐζουσιν. "If it were not Dionysus for whom they march in procession and chant the hymn to the phallus, their action would be most shameless. But Hades and Dionysus are the same, him for whom they rave and celebrate Lenaia." 58 The key to this fragment is in its word play. 59 Αἰδοῖα is connected with αἰδώς, which means "shame" but also "reverence." Thus, in this case, the word αἰδοῖα denotes the sacred phallic symbols, even if this word also occurs with the denotation of pudenda. Ἀναιδέστατα is to be intended as the "most shameless actions," but at the same time as αν-(῾)Άιδησ-τατα. Their actions would be "without Hades" if they were not performed for Dionysus. However, Dionysus and Hades are the same. The verbal connections between αἰδοῖα -αἰδώς -ἀναιδέστατα aim to underline the intimate equivalence of Dionysus-the god of vitality through madness-with Ἅιδης, the god of the underworld, we thus find the equivalence of the two opposites, life and death, in line with the Heraclitean concept of the structure of reality, which consists in the unity of opposites. 60 In the quoted fragments of Pherecydes and Heraclitus, the polyonymy is recorded, if not as an exception, at least as a particularity: the concept of the adherence of the names to reality is so strong, that it can be assumed that a plurality of names may correspond to a plurality of functions/attributes/domains, but the object is actually only one. Asymmetry is thus justified de facto as only seeming.
Heraclitus introduces us to the field of speculative thought. The ontological bond between names and objects is well attested by some other fragments, such as fr. 123 Kahn: 61 Ὁ θεὸς ἡµέρη εὐφρόνη, χειµὼν θέρος, πόλεµος εἰρήνη, κόρος λιµός. Ἀλλοιοῦται δὲ ὅκωσπερ ὁκόταν συµµιγῇ θυώµασιν ὀνοµάζεται καθ' ἡδονὴν ἑκάστου. 62 "The god: day night, winter summer, war peace, satiety hunger. It alters, as when mingled with perfumes, it gets named according to the fragrance of each one." The fragment is the subject of much discussion amongst scholars. 63 Some of them, such as Kahn and Bollack, and Wismann, maintain the text as attested by Hippolytus (Refut. 9.10.8). However, the majority prefer to follow the correction made by Diels who adds the word πῦρ between ὅκωσπερ and ὁκόταν. 64 Marcovich has added an ulterior correction: πῦρ ὃ. If this integration is to be accepted, the subject of ὀνοµάζεται is fire (πῦρ), if not, the subject is god (θεός).The meaning of καθ' ἡδονὴν ἑκάστου is also under discussion, either as "according to the pleasure of each one" or as "according to the fragrance of each θύωµα," which is the one we prefer. As has been pointed out by Diano and Serra, 65 following Kirk, and Marcovich, 66 Heraclitus affirms a real and substantial communion between names and the objects they denote: just as night and day coincide, god is night and day, not as a subject is its predicates, but as different substances are the same and only substance. The one-to-one relationship seems to be confirmed-and the asymmetry relegated to the realm of appearance-in the second part of the fragment, in which the god or the fire changes when mingled with a specific θύωµα, and its name consequently changes in order to designate each single mixture.
Name and function, conceived as essential and inseparable components of each object, 67 are equivalent in the well-known fragment 22 B 48 DK, 68 which registers a particularly complex example of polyōnymia: Τῷ οὖν τόξῳ ὄνοµα βίος, ἔργον δὲ θάνατος. "Then, the name of the bow is life; its work is death." here with the polysemic value of this word. Considered in its two possible pronunciations, it means 'bow' (βιός) and 'life' (βίος), and therefore is a contrast in itself: the result of the bow's action is in fact death. Moreover, at Heraclitus' time, the accents of words were not written, and consequently the difference between βιός and βίος would not have been immediately perceivable to the reader. 69 Therefore, this fragment records the synonymy/polysemy of τόξον/βιός and the polysemy/homonymy of βιός/βίος. The object has a double name, but only one effect on reality, which is death. As Robinson has pointed out "the fragment serves also as a striking instance of how names can indicate the reality (or an aspect of the reality) of a thing" (see fragments 23,32,67). 70 For Heraclitus all things are opposite to each other, creating continuous changes which find their composition in the harmony of the λόγος. Ordinary people can only see the opposites, while the sage is the one able to understand the superior harmony and unity which is beyond them: θεός and βιός are in fact names representing the harmony beyond the opposites. For this reason, even if Heraclitus conceives the word-object relation as one-to-one, he also remarks that (fr. 1 Kahn): Τοῦ δὲ λόγου τοῦδ' ἐόντος αἰεὶ ἀξύνετοι γίνονται ἄνθρωποι καὶ πρόσθεν ἢ ἀκοῦσαι καὶ ἀκούσαντες τὸ πρῶτον• γινοµένων γὰρ πάντων κατὰ τὸν λόγον τόνδε ἀπείροισιν ἐοίκασι, πειρώµενοι καὶ ἐπέων καὶ ἔργων τοιουτέων ὁκοίων ἐγὼ διηγεῦµαι κατὰ φύσιν διαιρέων ἕκαστον καὶ φράζων ὅκως ἔχει [...]. "Human beings are (always) unable to understand this λόγος which (always) is, both before they have heard about it and while they are hearing about it for the first time. They seem inexperienced in all the things happening according to this λόγος, even if they have experience of the words and the actions such as those I describe according to nature, distinguishing each one and showing how it is [...]." In this earlier phase of the pre-socratic age, asymmetry was clearly considered a deviation from the intuitively perceived as natural, one-to-one word-object relation, and this led several authors and thinkers to try and justify it, especially by means of etymology. In this age, etymology consists de facto in the search for the thing in the name, and cases of asymmetry, such as the existence of two names for a single deity, must be explained and justified. Thus, we can reasonably conclude that the first answer regarding the nature of this phenomenon is that asymmetry is only seeming. In this frame, Heraclitus can be considered as both archaic and as a starting point. Archaic, because he only conceives the word/underlying ἔργον (the thing) relation and not the one between the linguistic sign and the concept or idea. 71 However, his work is also an important starting point, since the problem as to whether the word-object relation is naturally correct or a mere convention actually has its origin in the afore-mentioned search for the "thing" in names, and mostly from the observation of the ceaseless changes of perceivable phenomena. 72 approach, even though the same work also provides a lengthy commentary on the two glosses, quoted as our starting point on the naturally established one-to-one relationship of signification. The impermanent nature of the objects which are temporarily considered as wholes seems to be emphasized, for instance in M 1.1 ll. 6-7: atha gaur ity atra kaḥ śabdaḥ. kiṃ yat tat sāsnālāṅgūlakakudakhuraviṣāṇy artharūpam sa śabdaḥ, "Now, in gauḥ what is [to be considered] the word? Is it the visible appearance, that which consists in the object possessing dewlap, tail, hump, hoofs and horns? Is this the word?" Thus, the actually existing objects of language could merely be the constituent parts of the cow and not the whole cow. This interpretation is suggested by the partial coincidence of this perspective on words with a very famous (possibly Sarvastivādin Buddhist) passage, included in a contemporary text, that is the Milindapañha [Mil] 73 (second century BCE). The latter is a work which claims to record a discussion between the king Milinda (supposedly representing the historical Indo-Greek King Menander) and the Buddhist monk Nāgasena. The matching passage questions the real entity of the object denoted by common words such as "chariot": (Mil 27) nāhaṃ bhante nāgasena musā bhaṇāmi, īsañ ca paṭicca akkhañ ca paṭicca cakkāni ca paṭicca rathapañjarañca paṭicca rathadạ ͂ ḍakañ ca paṭicca ratho 'ti saṅkhā samaññā paññatti vohāro nāmamattaṃ pavattatī 'ti, "Revered Nāgasena, I am not telling a lie: it depends on the pole, on the axle, on the wheels, on the body of a chariot, on the flag-staff of a chariot, on the yoke, on the reins, and on the goad, if 'chariot' exists as a denotation, appellation, designation, as a current usage, as a name." In fact, as Buddhists unanimously agree, from the ontological point of view, there is no whole independent of its parts, nor indeed do wholes exist at all. Nonetheless, elsewhere (M 1.220 ll. 22-4 ad Vt 10 ad A 1.2. 45 Vt 10) Patañjali reflects on the whole-part relationship and even employs the chariot example but in a different way. Indeed, he maintains that it is the whole and undivided word that guarantees the denotation and function of the parts, precisely because, only a whole chariot is fit for movement, while its constituents, if they are taken apart one by one (rathāṅgāni vihṛtāni pratyekaṃ), are not suitable for this purpose (vrajikriyāṃ praty asamarthāni bhavanti). The linguistic aim is to explain that the combinations (samudāya) of sounds have meaning, whereas the parts do not.
Indeed, this is another way of underlining the intrinsic asymmetry of language, since the autonomous identity of the whole is underlined as a reality which surpasses the sum of its parts. In our example, the chariot is something more than the sum of its parts, which by contrast would be useless if they were separated from each other. Moreover, if we try to proceed by subtraction by starting from the whole, we notice that something which has undergone a change with regard to one of its parts is by no means something else-as a consequence of this change. 74 Patañjali's mundane example is the case of the classification of animals: M I.136 ll. 9-10 ad A 1.1.56 Vt 10: tad yathā | śvā karṇe vā pucche vā chinne śvaiva bhavati nāśvo na gardabha iti, "When a dog has an ear or tail cut off, it remains a dog indeed. It does not become a horse or a donkey." Finally, the ontological solidarity between the whole and its parts seems to be insisted upon in the M, with regard to the connection between the linguistic unit and its segments or sub- units. One may recognize parts in the dravyas, yet these parts are still intrinsically integrated into the whole, as stated by means of the formula vṛkṣaḥ pracalan sahāvayavaiḥ pracalati, "A tree when it shakes, shakes with its parts," which occurs thrice in this work. 75 Wholes are the permanent frames upon which symmetry is re-established. On the other hand, this pattern opens the way to the well-known difficulty relating to the permanence of words. If words were to lose or modify some of their parts, like a dog losing its ear or tail, they would not be considered as permanent. In fact, grammarians elsewhere tend to resort to the substitution of full words (sarvapadādeśa), in order to avoid what could otherwise appear as "a change, a modification of words." 76 A word denoting a whole depends on a convention and warrants the mutual comprehension between speakers in everyday usage, but this kind of word prevents man from grasping the right knowledge.
It seems indisputable that the same question found in the Mil was in the background of Patañjali's reflections on this matter. The relevant examples are closely similar, but the answers are neatly different. For the Buddhist sources, whose pivotal doctrine is the denial of the ontological existence of individual existence, phenomenological individuality is merely admitted as being perceived illusorily, because of temporary combinations (aggregates) of physical and psychic, ultimate and indivisible constituents (dharmas) which are metaphysically a given datum. As a consequence, the impermanent aggregates which illude the human mind are only assumed to exist as names (nāmadheyamātra). The objects of denotation and the relevant words which denote them are not symmetric, and this asymmetry is tolerated in order to grant conventional communication, i.e. the common mutual comprehension of everyday life.
This conventional use of language is therefore often defended even though it is a recognized obstacle for the true knowledge. For instance, Buddha forbids any transgression of the limits of convention, based on some dialectical usage, as shown in Majjhima Nikāya [MN] 77 3.230 (janapadaniruttiṁ nābhiniveseyya, samaññaṁ nātidhāveyyāti): he is afraid that this can determine verbal incomprehensions or useless disputes, when for example the same object is denoted by different nouns. 78 The relevant example in MN 3.234-235 is the almost synonymous series of pāli words pāti, patta, vittha, sarāva, dhāropa, poṇa, pisīla, used in different parts of the country to denote a "bowl," but more properly matching with a number of different shapes, such as a vessel, a bowl, a cup, a goblet, etc. 79 Buddha is thoroughly persuaded that there are even words to which no object actually corresponds, such as pāli attan "self" or aham "I," as might be expected, but linguistic convention warrants a mutual comprehension. The example of milk, which changes into curd, butter, and clarified butter and is denoted with a different noun at each stage, shows how the continued use of the same noun khīra "milk" instead of the conventional noun denoting the specific states (dadhi, navanīta, sappi) is of no help (see Dīgha Nikāya 80 1.201). However, none of these nounsthe first one included-actually matches a given existing entity which can be known as such.
In the same perspective of the signification, another point which is shared by the Mahābhāṣya and by the Pāli canon has been highlighted by Bronkhorst in 1987. Throughout the long history of the Buddhist dharma theory, only a limited number of dharmas came to be accepted as truly self-existing entities 81 and, as is well known, according to the majority of Buddhist schools, their existence is merely momentary. Precisely in order to avoid the judgement of momentariness for sounds, words, and sentences, the Sarvāstivādin School postulatedprobably for the first time-self-existence for two linguistic dharmas, namely the vyañjanakāya and the nāmakāya/padakāya (sound and word). Patañjali also considers word and sound as the only two self-existing autonomous linguistic entities. 82 The sole morphological unit which Patañjali consistently considered self-existent is the inflected pada, whose autonomy is never questioned. The different treatment of morphological entities, described in the grammar as somehow originating dependently, reveals the illusory autonomy of sub-units of words-with the regular highlighting of the dependence-relationship between morpheme and morpheme in an inflected word or between an inflected word and another one in a compound. Therefore, the specific linguistic and speculative reflections on the conventional relation between word-forms and their objects, briefly exemplified here, were certainly broadly circulating in India in about the second century BCE. As a result, an explicit adhesion or confutation in the majority of the subsequent technical and speculative traditions could have been requested. Nonetheless, we wonder whether a comparable conventional explanation of the asymmetry of language was also advanced earlier, because the relevant Buddhist sources date back to at least the second century BCE, although it is possible that they constituted some portions of Buddha's preaching (thus dating back to sixth-fifth century BCE).
Indeed, in the Chāndogya-Upaniṣad [ChUp], 83 which possibly dates back to the sixth century BCE, language is in fact presented as a human undertaking that creates specific distinguished objects and fashions reality into its illusory discrete entities: The red visible appearance of fire is indeed the visible appearance of energy; the white, that of water; the black, that of food. The individuality of fire disappears. The specific modified form is a verbal undertaking, or better a name. 86 The perceptible reality is just this: the three visible appearances. The red visible appearance of the sun is indeed the visible appearance of energy; the white, that of water; the black, that of food. The individuality of the sun disappears. The specific modified form is a verbal undertaking, or better a name. The perceptible reality is just this: 'They are just the three visible shapes.'" The sapiential and substantially gnostic context of these passages aims at going beyond the discriminatory knowledge (viveka-jñāna) which is a condemned fruit of ignorance. In fact, it illusorily generates the plurality of perceptible appearances, which all prevent the common people from catching the ultimate oneness of truth (more precisely of that which is permanent, namely that which is ontologically/metaphysically really existent). Therefore, differently from the Mil passage quoted above, the aim here is to show that the shapes and names (rūpa and nāman) of things cannot be real, because the imperceptible one is the only substance that exists. The three basic evolutes mentioned-termed here as energy, water, and food-especially because they explicitly match the colours of red, white, and black respectively, might be considered as a sort of antecedent of the well-known three properties (guṇas) of the Saṃkhya tradition-the three qualities of primeval, irreducible, and immanifest matter (nature). The temporary individualities which language provides with labels/names depend on the different mingling of these three properties, but the only ones that actually (ontologically) exist are these three basic properties of nature.
The shared point is that if language did not provide these temporary combinations with proper designations, they would not be perceived at all. The multiplicity of commonly used

In Old Greek Sources
In the Greek culture, Parmenides is the first to explicitly declare that names have been imposed by humans (fr. 28 B 8 DK, ll. 38-40). Thus, we suppose that he does not believe that language merely arises by nature.
Τῷ πάντ' ὄνοµ(α) ἔσται, ὅσσα βροτοὶ κατέθεντο πεποιθότες εἶναι ἀληθῆ, γίγνεσθαί τε καὶ ὄλλυσθαι, εἶναί τε καὶ οὐχί. "The being will have as many names, as those which have been imposed by the mortals, who are convinced that they are true: 'to be born,' 'to die,' 'to be' and 'not to be.' According to Parmenides, the being is immobile, eternal, one and whole. However, human beings are deceived by opinion (δόξα), which makes reality appear as divided into different components, so that humans have imposed names in order to identify each of them (fr.

B 19 DK):
Ὅὕτω τοι κατὰ δόξαν ἔφυ τάδε καί νυν ἔασι καὶ µετέπειτ' ἀπὸ τοῦδε τελευτήσουσι τραφέντα· τοῖς δ' ὄνοµ' ἄνθρωποι κατέθεντ' ἐπίσηµον ἑκάστῳ. "In this way, according to opinion, these things are born and now are, then, once they have grown, they will die: humans imposed names as distinctive signs of each one." Unlike Heraclitus, for whom the ὀνόµατα are only nouns, Parmenides evidently thinks that they are also verbs and locutions. Both philosophers think that humans are deceived by the seeming multiplicity of reality, but only Parmenides explicitly states that they have imposed names on each part of it, so that it is clear that he conceives these names as a mere convention. According to this conception, asymmetry, which is once again identified with polyōnymia, seems to be a consequence of the deceitful knowledge of reality.
As far as Empedocles is concerned, the structure of the world is made up of four simple, eternal, and unalterable elements-earth, fire, air, and water-which are brought into union by Love and into separation by Strife (fr. 31 B 8 DK): […] Φύσις οὐδενὸς ἔστιν ἁπάντων θνητῶν, οὐδέ τις οὐλοµένου θανάτοιο τελευτή, ἀλλὰ µόνον µίξις τε διάλλαξίς τε µιγέντων This fragment clearly shows that nominating is a human prerogative. 87 Fr. 31 B 9 DK, ll. 3-5 (= fr. 56 Bollack) is even clearer about the matter. When the elements are mixed in the form of man, animals, plants or birds [...] τότε µὲν τὸ <λέγουσι> (?) 88 γενέσθαι, εὖτε δ' ἀποκρινθῶσι, τὸ δ' αὖ δυσδαίµονα πότµον· ἣ <γε> θέµις, 89 καλέουσι, νόµῳ δ' ἐπίφηµι καὶ αὐτός. "[…] then they call it 'to be born,' but when they separate, they call it 'unfortunate fate'; surely, they call it with good reason, and I myself agree with this custom." While Parmenides condemns names as manifesting the lack of understanding of reality, Empedocles accepts them as a useful convention for communication between humans, provided that they recognize the true structure of the world lying beyond them. In his opinion, even if reality is made up of momentary aggregations, humans can communicate with each other by giving names to each group. The norm provides for the reparation of a natural asymmetry between the names-which are stable and fixed by a convention-and the object they refer towhich is only seemingly unitary.
According to Proclus, Democritus explicitly discusses the arbitrariness of names by means of four arguments (fr. 68 B 26 DK):

4.
Ἐκ δὲ τῆς τῶν ὁµοίων ἐλλείψεως• διὰ τί ἀπὸ µὲν τῆς φρονήσεως λέγοµεν φρονεῖν, ἀπὸ δὲ τῆς δικαιοσύνης οὐκέτι παρονοµάζοµεν; τύχῃ ἄρα καὶ οὐ φύσει τὰ ὀνόµατα. "The argument from the lack of corresponding names: why do we derive 'being wise' from 'wisdom,' but from 'justice' we have no verb? Hence names exist by chance and not by nature." 92 Democritus thus sheds light on the lack of a perfect correspondence between language and reality. 93 According to the sources, the sophist Protagoras considers both the theories of Democritus and the doctrines about the ceaseless changes of the universe, expecially the one proposed by Heraclitus. 94 The latter doctrine surely underlies the Protagorean theory of relativism, according to which "of all things the measure is man, of things that are that they are, of things that are not that they are not" (Pl. Tht. 152a. Tr. Graham). If man is the measure, everything he experiences or says is true: as a consequence, two opposed discourses on the same matter are equally true. Plato and Aristotle link Protagoras to the eristic for his attempt to refute the speeches of his antagonists by exploiting the ambiguities of words. 95 He also treats the lack of correspondence which may exist between the 'natural' and the grammatical gender of names, blaming, for example, the use of names such as µῆνις 'anger' or πήληξ 'armour' as feminine rather than as masculine. 96 The scholars agree on the fact that Aristophanes puts this kind of Protagoraean research to shame in his Clouds, where Socrates notes that both the rooster and the hen are called with the masculine ἀλεκτρυών in Greek. It is, thus, necessary to coin a new feminine name, such as ἀλεκτρύαινα! Moreover, even though the name of the kneading-trough, ἡ κάρδοπος, is feminine, it belongs to the o-stem declension which is usually assigned to masculine names (658-680). An echo of the Protagorean research was also seen in the so-called 'battle of the prologues' in Frogs 1119-1197, 97 where it is however possible that in lines 1182-1195, Aristophanes is alluding to Prodicus' 'correctness of names' theory. 98 In fact, Prodicus, said by the Suda to be the "disciple of Protagoras," 99 deals with the socalled ὀρθότης τῶν ὀνοµάτων. 100 He was interested in etymology 101 and in homonymy, 102 even if most sources inform us about his research on synonymy. This is where Prodicus shows that words commonly considered as having the same object are not completely interchangeable. Plato's Protag. 337a-c might show Prodicus' method, which consists in considering two synonyms and in explicating the reason why they are actually (even if sometimes only slightly) different: εὐδοκιµέω is not the same as ἐπαινέω because the former is used to denotate praise bestowed without any deceit, while the latter should also indicate insincere praise.
We agree with Mayhew's opinion that Prodicus probably thinks that names are "stipulated (and so in an important sense conventional), but that it helped if they could be derived from or connected to the nature of what they name." 103 Protagoras and Prodicus are a step ahead of Empedocles. He accepts the conventionality of names as a useful means for communication (νόµῳ δ' ἐπίφηµι καὶ αὐτός), while the two sophists criticize the excessively free use of words, which should adhere to the denoted object with precision. Playing with ambiguities is in fact the best way to deceive, but also to be deceived.
One of the most famous sophists, Gorgias, sheds light on the impossibility for words to convey knowledge of things, because they are of a completely different nature (fr. 82 B 3 DK, § § 83-87): [...] ᾧ γὰρ µηνύοµεν, ἔστι λόγος, λόγος δὲ οὐκ ἔστι τὰ ὑποκείµενα καὶ ὄντα· οὐκ ἄρα τὰ ὄντα µηνύοµεν τοῖς πέλας ἀλλὰ λόγον, ὃς ἕτερός ἐστι τῶν ὑποκειµένον [...] Εἰ γὰρ καὶ ὐπόκειται [...] ὁ λόγος, ἀλλὰ διαφέρει τῶν λοιπῶν ὑποκειµένων, καὶ πλείστῳ διενήνοχε τὰ ὁρατὰ σώµατα τῶν λόγων• δι' ἑτέρου γὰρ ὀργάνου ληπτόν ἐστι τὸ ὁρατὸν καὶ δι' ἄλλου ὁ λόγος [...] Τοιούτων οὖν παρὰ τῷ Γοργίᾳ ἠπορηµένων οἴχεται ὅσον ἐπ' αὐτοῖς τὸ τῆς ἀληθείας κριτήριον. "[…] That by which we communicate is speech, but speech is not the subsisting and existing things themselves. Therefore we do not communicate to our neighbours the existent things, but speech, which is different from the subsisting things […] Even if speech does subsist, it differs from other subsisting things, and visible bodies differ most markedly from words. For the object of sight is grasped by a different organ than speech […] Since such things are called into question by Gorgias, as far as they are concerned the standard of truth fails." 104 The awareness of the instability of the perceivable phenomena, which had just begun with Heraclitus, brings thinkers, such as Parmenides and Empedocles, to recognize the conventional status of words and the instability of the word-object relation. Gorgias develops this assumption by asserting that it is not possible to know reality by means of words. In this second phase of pre-socratic thought, the word-object relation is thus considered as a mere convention. A well-established and naturally one-to-one correspondence between word and object does not exist, and this is also the reason why asymmetry exists with no exception. 105

Conclusions
The comparative analysis of Indian and Greek traditions clearly shows that both followed the same iter which led from the intuitive belief in a natural one-to-one word-object relation to the conviction that this relation is a mere convention and a human undertaking. This conclusion is strictly bound to the observation of the ceaseless change and instability of reality. Thus, asymmetry is at first considered as only seeming, and the role of language as a means of knowledge is preserved; subsequently, the existence of asymmetry is recognized and accepted as part of the convention on which language is based. The consciousness of the lack of a wellestablished and natural word-object relation led to the conclusion that language is definitively an unreliable means of knowledge.
In both India and in Greece, paretymologies play the role of checking the one-to-one word-object relation. This is mostly applied to divine names, as shown by the passages from Ṛgveda, Atharvaveda and the Nirukta, and in those by Homer, Hesiod, Pherecydes of Syros, and Heraclitus.
In both the selected groups of sources, the reflection about asymmetry encompasses the thought on the almost technical derivation of words. Compelling evidence of this is offered by some examples from the Nirukta, on the one hand, and from Democritus, on the other.
In the passages from the Vedic Upaniṣads and the Buddhist sources, language is presented as a human undertaking, which determines the deception of the ignorant. The latter erroneously concentrates on whole objects and their matching nouns instead of on their single perceptible parts and is prevented both from perceiving that which really exists beyond phenomena and from having access to the phenomena as such (dharmas). This kind of speculation, particularly the Upaniṣadic one, perfectly matches the thought of Parmenides and Heraclitus, who are convinced that the majority of people are not able to comprehend reality beyond the multiplicity of phenomena on which names are usually imposed. The Buddhist sources tolerate the asymmetry between the objects of the denotation and the relevant words, exactly as Empedocles does, in order to warrant the mutual comprehension of speakers.
The most important difference between the two cultures, as emerges from these documents, is that a technical-grammatical reflection on language was developed earlier on in India. Therefore, the watershed in the development of Indian reasoning on word-signification is the work of genuine linguists, such as Pāṇini (fourth century BCE) and his first commentators (third-second century BCE). On the contrary, as far as Greek sources are concerned, we are forced to merely adopt the general distinction between thought before and after Socrates.
Nonetheless, both in the ancient Greek sources and in the most ancient Indian ones such as the Vedic passages here quoted, in the period considered here (respectively, mid eighth century BCE-fifth century BCE and twelfth century BCE-fifth century BCE), language is not the specific object of interest and those who reflect on it are often not grammarians. As a consequence, linguistic phenomena are at first occasionally observed in literary texts, and only later do they become part of technical traditions and of wider philosophical theories about the structure of the world. It is also noteworthy that it is only in India that linguistic theories are mostly influenced by the need to preserve the validity of the ritual 106 and the permanent relation between the words of the sacred text and the objects they refer to.
It is reasonable to assume that the abovementioned specific need is also sufficient justification for the frequent primacy of speculations on language in the subsequent scientific and technical Indian scenario. It is documented at a very early date in the Nighaṇṭu's exegetical-lexical lists (commented on by the Nirukta), which, as is well-known, are mainly arranged according to the categories of synonyms, homonyms and theonyms, thus according to the asymmetries of language. This system seems to presuppose the crucial concept of substitution, 107 which may already have played a decisive role in the ancient Vedic ritual-poetic contest in assuring preeminence in the sacrificial arena and as a consequence in society. Later the substitution considered as "a theory of truth" (Kahrs 1998: 173) became the focus of the majority of philosophical assertions in Upaniṣadic thought, by means of the so-called "equivalences," which are often pondered substitutions of objects or concepts with others considered equivalent. To a large extent, this system of "replacement" could be considered as a peculiarly Indian and ritually oriented development of the wide potentialities of asymmetry.
Instead, in Greece the problem of asymmetry was almost always strictly linked to that of the reliability of the words and of the λόγοι they make up. This relation was historically decisive both in the field of rhetoric intended as art of persuasion, especially in political and legal contexts, and in that branch of philosophy which investigates the possibility of knowing reality by means of logical reasoning. 108