Paninian And Veda Reconsidered

Author(s): Johannes Bronkhorst
Publisher: Johannes Bronkhorst
Catalog link: https://jainqq.org/explore/269665/1

Page #1

PANINI AND THE VEDA RECONSIDERED
Johannes Bronkhorst
The relationship between Panini and the Veda has been much debated. The presupposition underlying a major part of this debate has been that much or even most of Vedic literature existed in its present form prior to Panini. In this article an attempt will be made to establish, as far as possible, the relationship between Panini and the Veda without taking the correctness of this presupposition for
granted.
1.1 A fundamental question is whether Papini knew the Vedic texts, i.e., the ones with which he was familiar, in the same form as we do. Were the Vedic texts that Pāņini knew identical in all details with the editions we have now? It appears that the answer to this question must be negative.
It is not always possible to decide that a text has not reached us in its original form. In the case of metrical texts this may be possible, however, and to some extent we may be in a position to determine what the original text was like. This is the case regarding the Ṛgveda. In another study (Bronkhorst 1981) it has been shown that certain rules of sandhi of the Aṣṭādhyāyi fit an earlier stage of the text of the Ṛgveda than the one we now have. The conclusion was
75

Page #2

drawn that "the lack of agreement between the Aṣṭādhyāyi and our Ṛgveda may henceforth have to be looked at through different eyes. Certainly, where phonetic questions are concerned, Panini may describe an earlier form of the Ṛgveda, and may not deserve to be blamed for being lacunary..." (pp. 91-92).
This conclusion has far-reaching implications. The Ṛgueda has been handed down with great care, with greater care perhaps than any other Vedic text. Yet even here Panini's rules of sandhi do not fully agree with the present text, although we know that at least some of them once fitted. How much less can we expect full agreement between Panini's rules of sandhi and all other Vedic texts. This means that a comparison of Panini's rules of sandhi and the Vedic evidence, if it is to be made at all, must be made with the greatest care. A straight confrontation of Panini's rules with the Vedic facts cannot be expected to yield more than partial agreement, and says little about the state of affairs in Panini's day. In the present context it is important to recall that "Panini's rules on Vedic sandhi do not necessarily describe the sandhi which was actually used in the Vedic texts which Panini had before him. Rather, they describe the sandhi as it ought to be according to Panini. This is confirmed by the circumstance that Panini sometimes gives the opinions of others besides his own, for example, in P.8.3.17-19" (Bronkhorst 1982, 275).2
A development in tone patterns, too, must have taken place after Panini. Kiparsky (1982, 73) sums up the results of an investigation into this matter: "[T]he tone pattern described by Panini represents an older stage than that
76
described for the Vedic samhitas by the Prätiśäkhyas. While the samhitäs themselves are of course older than Panini's grammar [?; see below], we may assume that they were accented in Panini's time with the tone pattern described in the Aṣṭādhyayi, and that their present tone pattern, as well as the Prätisśākhyas that codify it, are post-Paninian revisions." It is true that Kiparsky derives the different tone patterns from accent properties belonging to morphemes that are stable in time. Yet it is at least conceivable, also, that these accent properties changed in the time before the tone patterns reached their final form. This means that little can be concluded from such deviations from Panini in the accentuation of Vedic words as occur in arya (Thieme 1938, 91f.; Balasubrahmanyam 1964; 1969), hayana (Balasubrahmanyam 1966), jyestha and kanistha (Devasthali 1967, 7-8), arpita and justa (Balasubrahmanyam 1974), śriyase (Balasubrahmanyam 1969; 1972), voḍhave (Balasubrahmanyam 1983), and vrsti, bhūti, and vitti (Keith 1936, 736)."
3
This is further supported by the fact that accents were not noted down until very late (Thieme 1935, 120f., 129f.). A passage in the Satapatha Brahmana (1.6.3.10) gives further proof for this. There Tvast pronounces a mantra wrongly, and as a result Vrtra is killed by Indra instead of the reverse. The mantra concerned is indrasatrur vardhasva. The later tradition--Patanjali's Mahābhāṣya (1:2, 1. 12), Paniniya Siksä (verse 52), Bhaṭṭabhaskara and Sāyaṇa (on TS 2.5.2), etc.--agrees that the mistake concerned the accent: an intended Tatpuruşa compound 'killer of Indra' becomes a Bahuvrihi 'whose killer is Indra'. The formulation of TS 2.5.2.1-2--yad abravit svahendraśatrur vardhasveti tasmād
77

Page #3

asyendraḥ satrur abhavat--fully agrees with this. MS 2.4.3 is even clearer: svahendrasatrur vardhasva itindrasyahainam satrum acikirşad indram asya satrum akarot. Yet the Satapatha Brahmana formulates the story in a way that can only be explained on the assumption that there was no way to make the difference in accentuation visible. Rather than writing (or reciting!) the Tatpurusa compound with the appropriate accent, it analyzes the compound into indrasya satruḥ. The passage then reads: atha yad abravid indrasatrur vardhasveti tasmad u hainam indra eva jaghāna / atha yaddha sasvad avakṣyad indrasya satrur vardhasveti sasvad u ha sa indram evahaniṣyat 1.
These considerations show that any comparison between the linguistic data in Panini and those in the Veda must be extremely careful in the fields of sandhi and accentuation. They also suggest that in other respects the Vedic texts known to Panini may have undergone modification since Pāņini's time.
8
As an example of a feature that may have changed since Panini, consider the word ratri/ratri in the mantras of the Taittiriya Samhita. According to P.4.1.31 (ratres cajasau), rātrī occurs in ritual literature (chandasi, see below) before all endings except the nominative plural (cf. Bhat 1968; Wackernagel 1896-1930, 3:185f.). Five times the mantras of the Taittiriya Samhita contain the word in a form that allows us to determine whether rätri or ratri is used. Twice (TS 4.3.11.3 and 5.7.2.1) it is rātri, thrice ratri. However, it is not impossible that originally all five occurrences had a form of ratri. TS 4.1.10.1 (rätrim rātrim aprayāvam bharantaḥ) recurs as rātrīm rātrīm (at MS 2.7.7
78
and 3.1.9; KS 16.7 and 19.10; and SB 6.6.4.1). TS 4.4.1.1 (ratrim jinvośigasi) occurs as rätrim jinvo at KS 17.7. In these two cases the shortening of i to i was a minor change. More problematic seems to be TS 7.4.18.1 (rätrir āsīt pisangila), to which no parallels with long i correspond (Bloomfield 1906, 823). Here a substitution of ratri would lead to rātry äsit, which differs rather strongly from the mantra as we know it. However, no such objection can be raised against an earlier rätri āsīt; this in its turn might be looked upon as the result of sandhi applied to rātrī āsīt, by P.6.1.127 (iko' savarne sakalyasya hrasvas ca), a rule of sandhi that also held in the Ṛgveda, at least according to Sakalya (see Bronkhorst 1982a, 181).
1.2 The second introductory question we have to ask is whether or not Panini's Vedic rules were meant to be universally valid in the Vedic texts. Our observations on sandhi have made it clear that here, at least, there is nothing to contradict the supposition that Panini's rules were meant to be adhered to throughout. (This does not necessarily mean, however, that the texts known to Papini always had Panini's kind of sandhi.) It is at least conceivable that all the Vedic rules of the Astädhyayi were meant to be strictly followed unless the opposite is explicitly stated.
This takes us to the main point of this subsection. If Panini's Vedic rules were not meant to be followed strictly, this should have been indicated in the Aṣṭādhyāyi. Kiparsky (1980) has shown that Panini distinguished three kinds of optionality: va 'preferably', vibhāṣā 'preferably not', and anyatarasyam 'either way. This means that Pāņini used various means to indicate optionality. As a matter of fact,
79

Page #4

option is indicated in a number of Vedic rules. P.1.2.36, 6.2.164, and 7.4.44 read vibhāşa chandasi, P.1.4.9 (sasthiyuktas chandasi vä), P.8.3.49 (chandasi vä'prämreḍitayoḥ), P.5.3.13 (va ha ca chandasi), P.3.4.88 and 6.1.106 (vā chandasi), P.6.4.5 and 6.4.86 (chandasy ubhayatha), P.6.4.162 (vibhāṣarjoś chandasi), P.8.2.70 (amnarudharavar ity ubhayatha chandasi), P.8.3.104 (yajusy ekeṣām), P.8.3.119 (nivyabhibhyo'ḍ vyavāye va chandasi), P.8.3.8 (ubhayatharksu), and P.6.4.9 (vā ṣapūrvasya nigame). The words bahulam chandasi 'variously in ritual literature' occur no less than seventeen times together, not counting the rules wherein they may have to be continued. In P.1.2.61 (chandasi punarvasvor ekavacanam) and 62 (visakhayoś ca [chandasi]), the word anyatarasyām is in force from P.1.2.58, and is not cancelled until nityam in 1.2.63. In P.6.1.52 (khides chandasi) there is continuation of vibhāṣā from sutra 51, cancelled by nityam in 6.1.57. P.3.1.85 (vyatyayo bahulam) continues chandasi from 3.1.84 (chandasi sayaj api), which itself indicates optionality by means of the word api. Similar devices are used in P.1.4.81 (chandasi pare'pi), and 82 (vyavahitaś ca); P.3.3.130 (anyebhyo'pi drsyate [chandasi 129]); P.5.3.14 (itarabhyo'pi drsyante [chandasi 13]); P.6.3.137 (anyeṣām api drsyate [rci 133][?]); P.6.4.73 and 7.1.76 (chandasy api dṛśyate); P.7.1.38 (ktväpi chandasi); P.5.2.50 (that ca chandasi); P.5.3.20 (tayor därhilau ca chandasi); P.5.3.33 (pasca paścă ca chandasi); P.5.4.12 (amu ca chandasi); and P.5.4.41 (vykajyeṣṭhābhyām tiltatilau ca chandasi). P.3.2.106 (liṭaḥ kānaj vā) is confined to ritual literature because only there lit occurs (P.3.2.105 [chandasi lit). P.8.1.64 (vaivaveti ca chandasi) continues
80
vibhāşă (63), cancelled by nityam in 8.1.66. P.6.1.209 (justärpite ca chandasi) continues vibhāşä from 208, discontinued by 6.1.210 (nityam mantre). In P.6.3.108 (pathi ca chandasi) the word ca continues vibhāşă from 6.3.106 (cf. Kiparsky 1980, 62). P.8.3.105 (stutastomayoś chandasi) appears to continue ekeşam from 8.3.104. P.4.4.113 (srotaso vibhāṣa dyaḍḍyau) continues chandasi from 4.4.110.
Nityam in P.4.1.29 (nityam samjñāchandasoḥ), in 4.1.46 and 7.4.8 (nityam chandasi), and in 6.1.210 (nityam mantre), does not indicate that here, exceptionally, some Vedic rules are universally valid. Rather, it is meant to block the option that is valid in the preceding rules, as so often occurs in the Aṣṭādhyayi. We have no alternative but to assume that, just as in his other rules, Panini's Vedic rules not indicated as being optional were meant to be generally valid.10
From this we must conclude that deviations from Päņini in the Vedic texts known to Panini either did not exist in his time or were not considered correct by him.
1.3 We now come to the question of what range of literature Panini considered "Vedic" in one way or another. This is best approached by studying Panini's use of the word chandas by which he most often refers to Vedic literature. It is clear that Pāņini employs this word in a special way. The most common meaning of chandas is 'meter', and then 'metrical text'. But this is not the only sense in which Panini uses it. Thieme (1935, passim, esp. 67-69) showed that rules given under chandasi 'in chandas' are also valid for prose passages (brahmana and yajus). He therefore rendered chandasi as 'in Sacred Literature'. Thieme criticizes
81

Page #5

Liebich's (1891, 26) translation 'pre-classical language', saying: "I do not think it an appropriate translation, since it appears to endow Pāṇini with a historical perspective he hardly could have possessed" (p. 67).
This makes sense, but a major difficulty remains. Many of the forms taught under the heading chandasi occur in Sutra texts. Instances are numerous and only a few will be given here. The name Punarvasu, used optionally in the singular in chandas according to P.1.2.61 (chandasi punarvasvor ekavacanam (anyatarasyām 58]), is so found at Vişnusmyti (78.12) and Vass (1.5.1.5), besides several places in the Black Yajurveda. The singular of visakha, only allowed chandasi by P.1.2.62 (viśākhayos ca), occurs similarly at VaSS 2.2.2.14. The grammatical object of the root hrs can have an instrumental ending in chandas, according to P.2.3.3 (tytiyä сa hoś chandasi). One instance is MSS 1.6.1.23 (payasa juhoti dadhnā yavägvājyena vă (cf. Thieme 1935, 10]). Some forms are only attested in Sütras. Khanya(P.3.1.123) only occurs in LSS 8.2.4 and 5; (pra-)stävya(id.) in LSS 6.1.20; unniya (id) in Sacs 4.14.4; and yaśobhagina (P.4.4.132) in HiśS 2.5.43, 6.4.3.
It seems safe to conclude that Pānini's term chandas covered more than just 'Sacred Literature'. We may have to assume that certain works, primarily the ritual Sütras, and among those first of all the Srauta Sütras, belonged to a fringe area wherein Vedic usage was sometimes considered appropriate. The effect of this assumption for our investigation is that, where a chandas word prescribed by Panini is attested in one Vedic text and in one or more Sütras, we are
not entitled to conclude that Pāņini certainly knew that Vedic text
1.4 The final introductory question we have to consider is the following. Are Panini's Vedic rules descriptive or prescriptive? To be sure, to some extent they describe the language that Pāņini found in Vedic texts, and are therefore descriptive. But are they exclusively so? It may well be that Vedic texts were still being composed in Panini's day, and that he gives in his grammar guidelines regarding correct Vedic usage. This possibility has been discussed elsewhere (Bronkhorst 1982, 275f.) and is further strengthened by the evidence to be provided in the following sections of this article. Here attention may be drawn to another reason to conclude that at least some of Pāṇini's rules may have been meant to be prescriptive, besides, or rather than, being descriptive. They may have been composed with something like wha in mind.
Oha" is the term used to describe the adjustments Vedic mantras undergo to make them fit for other ritual contexts. An original mantra such as agnaye tvä justam nirvapāmi, directed to Agni, can become modified into süryaya tvä juştam nirvapāmi, directed to Sürya. Devir āpaḥ śuddha yüyam (MS 1.1.11, 1.2.16, 3.10.1; KS 3.6), directed to the waters, becomes deva ajya suddham tvam when directed to clarified butter (ājya). Sometimes only the number needs adjustment, as when ayur āśāste (MS 4.13.9; TS 2.6.9.7; TB 3.5.10.4) becomes āyur āśāsāte or ayur āśāsate. Only the gender is modified when jur asi dhytā manasa justā vienave tasyās te satyasavasaḥ (MS 1.2.4, 3.7.5; KS 2.5, 24.3; TS 1.2.4.1, 6.1.7.2; VS 4.17; SB 3.2.4.11; SBK 4.2.4.9) becomes
82

Page #6

14) specifies that which is not a mantra without mentioning tiha! Apparently, at one time, modified mantras were mantras.
This view is supported by the fact that modified mantras have actually been included in the Vedic collections as mantras. A particularly clear example is the long adhrigu passage that occurs, or is discussed, in MS 4.13.4, KS 16.21, TB 3.6.6, AB 2.6-7 (6.6-7), KB 10.4, ĀsvŚs 3.3, and SSS 5.17, with this difference: TB, AB, KB, and SŚs have medha patibhyam where MS and KS have medha pataye. Interestingly, the difference is explained in AB 2.6.6 (6.6.6) in the following words:
jur asi dhrto manasi justo vişnave tasya te satyasavasah because a bull is under discussion.
Another interesting question is whether modified mantras are in fact mantras themselves. The later Mimāmsă tradition appears to be unanimous in its opinion that they are not PMS 2.1.34 and Sabara's Bhäşya thereon state explicitly that the result of üha is not a mantra, and all later authorities in this field seem to have followed their example. This opinion is found, perhaps for the first time, in Apss 24.1.35, which reads anāmnātās tv amantra yatha pravarohanāmadheyagrahaņāniti "Die nicht (im Mantra. oder Brāhmanateile) überlieferten Teile sind indessen nicht als Mantra zu betrachten, z. B. der Pravara, die 'Verschiebung (üha), die Nennung eines Namens" (tr. Caland 1928a, 387).
It is not surprising that modified mantras were not considered mantras in their own right from an early date onward. After all, the opposite opinion would leave almost unlimited scope for creating new mantras. At a time when efforts had been made to gather all mantras into Vedic collections this must have been undesirable.
Yet there are clear traces of evidence that modified mantras were not always considered nonmantras. As late an author as Bharthari (fifth century A.D.)," who includes a long discussion on üha in his commentary on the Mahabhäşya (Ms 2b9 f.; AL 5.18 f; Sw 6.17 f.; CE Ähn. 1, 5.1 f.) mentions "others" who think that modified mantras are themselves mantras." And several Srauta Sutras make no mention of the nonmantric nature of modified mantras in contexts in which that would have been appropriate, for example, Bhass (6.15), MSS (5.2.9), and SSS (6.1). Moreover, HiśS (1.1.13
sa yady ekadevatyaḥ paśuḥ syān medha pataya iti brüyat yadi dvidevatyo medha patibhyām iti yadi bahudevatyo medha patibhya ity etad eva sthitam If the victim be for one deity, 'for the lord of the sacrifice' (medha pataye) he should say; if for two deities, 'for the two lords of the sacrifice' (medhapatibhyām]; if for many deities, 'for the lords of the sacrifice' (medha patibhyah). That is the rule. (Tr. Keith 1920, 138)
This is as clear a case of üha as is possible.
TS 2.3.10.1-2 repeats the same sacrificial formula four times, with differences in number, in a single passage in order to adjust it to different numbers of gods:
asvinoh prāno'si tasya te dattām yayoh prāno'si svāhā indrasya prāno'si tasya te dadātu yasya prāno'si svāhā

Page #7

miträvaruṇayoḥ prano'si tasya te dattam yayoḥ prano'si svāhā visveṣām devānām prāno'si tasya te dadatu yeṣām prāno'si svāhā
The question we must now consider is to what extent the Vedic rules of the Aṣṭadhyayi can be looked upon as having been composed with this kind of uha in mind. Obviously, it cannot be maintained that this was the only purpose of these Vedic rules, for some were undoubtedly intended to describe isolated Vedic facts. But this does not exclude the possibility that uha was one of the purposes for which some of the Vedic rules of the Aṣṭadhyayi were formulated.
There is some reason to accept this last view. Some Srauta Sutras lay down rules pertaining to the modification of certain verbal forms. MSS 5.2.9.6, for example, lists the following acceptable modified forms: adat, adatām, adan, ghasat, ghastam, ghasan, aghasat, aghastam, aghasan, karat, karatām, karan, agrabhit, agrabhiṣṭām, agrabhişuḥ, and akṣan. AsvSS 3.4.15, similarly, lists adat, ghasat, karat, juşatām, aghat, agrabhit and avivṛdhata. SSS 6.1.5, finally, lists adat, ädan, ghastu, ghasantu, aghasat, aghasan, or aghat, akşan, agrabhit, agrabhişuḥ, avivṛdhata, avivṛdhanta, and others. This shows that there was concern in ritual circles regarding the correct use of certain verbal forms in modified mantras. Among the recurring forms are the aorists of the roots ghas, ad," and kr.
16
The shared concern of AsvSS 3.4.15, SSS 6.1.5, and MSS 5.2.9.6 is explained by the fact that most of the modifications are meant for virtually identical texts, the so-called Praisa sūktas, in particular RV Khila 5.7.2 (f and 1), which
86
correspond to MS 4.13.7 (p. 208, 1.3-7) and 4.13.9 (p. 211,
1.5-12).
It is very probable that Panini knew the Praisa suktas in which these modifications were to take place, for Scheftelowitz (1919, 47f.) has adduced reasons to believe that the Praisas are among the oldest Vedic texts in prose. This allows us to surmise that a Paninian sutra may have been composed partly to solve this same problem. This sūtra would then be P.2.4.80 (mantre ghasahvaraṇaśavṛdahādvṛckygamijanibhyo leh), which deals with the aorists of a number of roots, among them ghas and kr, in a mantra. It favors here such forms as (a)ghat, (a)ghastam, akşan and akaḥ, and akran (not in all cases the same forms as the above Śrauta Sūtras).
If it can be accepted that P.2.4.80 was composed to serve the purpose of uha (besides other purposes), the same may be true of other rules of the Aṣṭādhyāyā. This, in turn, would mean that these rules not only describe Vedic data but also prescribe the means for modifying Vedic mantras when necessary. This implies that we cannot always be sure that Pănini's Vedic rules describe forms that occurred in Vedic texts known to Panini. Unattested forms accounted for by rules in the Astadhyāyi do not, then, in all cases have to have been part of texts that are now lost.
2. We can now turn to the main part of the present investigation: an attempt to determine which Vedic texts Pāņini knew and which he did not. The above considerations make it clear that in this context Panini's rules on sandhi and accent will be of little help. Moreover, none of the rules that concern details of the phonetic shape of words,
87

Page #8

i.e., the orthoepic diaskeuasis of texts, can be relied upon to determine which texts Pāṇini knew, for the simple reason that these features may have changed, and in some cases certainly changed, after him. Our enquiry must in the main rely on word-forms prescribed in the Aşțădhyāyi.
Here another consideration arises. We have decided to take Panini seriously, but this does not mean that we demand his grammar to be complete. Nor does it exclude the possibility that he made occasional mistakes. It does, however, imply that, where Panini ciearly and explicitly excludes certain features from the Vedic language, we must regard with suspicion the Vedic texts containing those features,
We proceed in a twofold manner. On one hand, we collect forms prescribed by Panini for Vedic and attested in but one Vedic text and nowhere else. If a sufficient number of such forms are found for a particular Vedic text and nothing else pleads against it, we may then assume that this Vedic text was known to Panini. On the other hand, we shall look for Vedic texts that contain features excluded by Panini. If the number of such features is sufficiently large, we may consider the possibility that Pāṇini did not know these texts. This double approach will provide us with the material to be evaluated in subsequent sections.
2.1 Many words prescribed by Panini are found only in the Rgveda. Some examples are vrkati (P.5.4.41) at RV 4.41.4; cicyuşe (P.6.1.36) at RV 4.30.22; yajadhvainam (P.7.1.43) at RV 8.2.37; jag?bhma (P.7.2.64) at RV 1.139.10 and 10.47.1;"' vrşanyati (P.7.4.36) at RV 9.5.6; tetikte (P.7.4.65) at RV 4.23.7; and svatavāmh payuh (P.8.3.11) at RV 4.2.6.
2.2 Three words prescribed by Panini for Vedic are only found in the Taittiriya Samhita: khanya- (P.3.1.123) at TS 7.4.13.1; the denominative kavya. (P.7.4.39) at TS 7.1.20.1; and anghuh (P.6.1.36) at TS 3.2.8.3. Note that all three words occur in mantras. Thieme (1935, 64) was of the opinion that a fourth word, brahmavadya (P.3.1.123), is found only in the Taittiriya Samhita. This word occurs in a brāhmana portion (at TS 2.5.8.3) but not only there; it is also found at JUB 3.2.3.2; ApSS 21.10.12; and VādhSs (Caland 1928, 176). Thus, no direct evidence remains that Panini knew the brāhmana portion of the Taittiriya Samhita.
2.3 Not all the evidence produced by Leopold von Schroeder (1879, 194f; 1881-86, 1:xi f., 2.viii f.) to show that Panini knew the Maiträyani Samhita can stand scrutiny. Some cases are derived not from Pānini but from his commentators. Others correspond to rules of Panini that are not confined to Vedic usage; these cases do not prove that Panini knew the Maitrāyani Samhitä, or a part thereof, for the simple reason that the words concerned were apparently also in use in other than ritual contexts. Finally, there are cases wherein Schroeder was mistaken in thinking that certain Vedic words prescribed by Panini occurred only in the Maiträyani Samhita and not in other texts.
However, the following cases can be used to establish Pāṇini's acquaintance with at least certain parts of the Maitrāyani Samhitā. P.3.1.42 teaches the Vedic (chandasi, but amantre) verbal forms abhyutsādayām akah, prajanayām akaḥ, and pāvayām kriyāt. They occur at MS 1.6.5, 1.6.10 and 1.8.5, and 2.1.3, respectively, and nowhere else. The Vedic (nigame) forms sādhyai and sādhva (P.6.3.113) are

Page #9

nowhere found except in MS 1.6.3 and 3.8.5, respectively. Agriya. (P.4.4.117) is only attested at MS 2.7.13, 2.9.5, and in the colophon to 3.1.10. Noncompounded bhavienu (P.3.2.138) is found only at MS 1.8.1. Praniya- (P.3.1.123) is found at MS 3.9.1 and nowhere else; ucchişya- occurs only at MS 3.9.2. Purişyavāhana (P.3.2.65) is found only at MS 2.7.4.
2.4 Vedic forms attested only in the Kathaka Samhita are the following (cf. Schroeder 1880; 1895): ramayām akah (P.3.1.42) at KS 7.7; upacayyaprda (P.3.1.123) at KS 11.1; and kşariti (P.7.2.34) at KS 12.11. One word occurs only in the Käthaka Samhita and in the Kapisthala Samhitä. Since the latter "is practically a variant of the Kathaka" (Gonda 1975, 327), it is here included: jagatya- (P.4.4.122) at KS 1.8 - Kaps 1.8, and at KS 31.7. Adhvarya in P.3.1.123 may indicate acquaintance with KS 35.7 = Kaps 48.9 (Thieme 1935, 23-24; Goto 1987, 191, n. 355).
2.5 A Vedic form found exclusively in the Atharvaveda is ailayit formed by P.3.1.51 (cf. Thieme 1935, 64); it occurs at AVS 6.16.3. Sivatāti (P.4.4.143) is only found at AVP 5.36.1-9. The word māmaki, formed by P.4.1.30, occurs only AVP 6.6.8.19
2.6 Two Vedic forms occur in the Latyāyana Srauta Sutra of the Samaveda and nowhere else (except, of course, in the later Drähyäyana Srauta Sūtra, which is often no more than a recast of the former): khânya- (P.3.1.123) at LSS 8.2.4 and 5 (DrSS 22.2.5 and 6); and (pra-)stāvya. (id.) at LSS 6.1.20 (DrŚS 16.1.22 and 18). Hvarita (P.7.2.33) occurs only in a mantra in MSS 2.5.4.24d and 4.4.39. Sanim sasanivamsam (P.7.2.69) occurs in mantras in MSS 1.3.4.2 and Vass
1.3.5.16 (cf. Hoffmann 1974). Dādharti is only attested in JB 2.37.20 Yasobhagina (P.4.4.132) is only attested Hiss 2.5.43 and 6.4.3.
3.1 We now turn to forms excluded by Panini.
P.3.1.35 (kaspratyayād am amantre liți) forbids a periphrastic perfect to occur in a mantra, yet AVS 18.2.27 has gamayām cakāra (cf. Whitney 1893, 249). AVP 18.65.10 has gamayām cakartha.
P.5.1.91 (vatsaräntäc chas chandasi) prescribes -iya after words ending in vatsara, resulting in forms like samvatsariya. The next rule, 5.1.92 (samparipūrvät kha ca), adds -ina in the same position, provided that vatsara.is preceded by sam- or pari. This means that Panini did not know, or approve of, forms wherein vatsarina is not preceded by sam- or pari.. Yet such forms occur: idāvatsarina at TB 1.4.10.2 and anuvatsarina at TB 1.4.10.3.
P.5.4.158 (tas chandasi) forbids the addition of kap after a Bahuvrihi compound ending in . An exception is brahmanabhartyka (AA 5.3.2).
P.6.3.84 (samānasya chandasy amūrdhaprabhytyudarkeşu) forbids substitution of sa- for samāna before mürdhan, prabhyti, and udarka. Yet this substitution has taken place in saprabhyti (PB 15.1.6 and KB 20.4, 21.4, etc.); sodarka (PB 13.7.9, 13.8.1, 13.8.4, and 13.8.5; and KB 20.4, 21.4, etc.).
P.7.1.26 (netarac chandasi) prohibits the use of neuter itarad in ritual literature. Yet it occurs at AB 6.15; KB 12.8; SB 4.5.8.14 and 13.8.2.9; TB 3.10.11.4; JB 1.213, 2.75, and 2.249; and at SadB 4.3.7, 4.4.10, and 4.5.8.
90

Page #10

P.7.2.88 (prathamāyāś ca dvivacane bhāṣāyām) prescribes the nominatives avām and yuvām with long penultimate a for secular language, thus excluding these nominatives from the Vedic language. Yet they occur in äväm (AB 4.8; Sana 5.7; SB 4.1.5.16 and 14.1.1.23; BAU[K] 3.2.13; CHU 8.8.1) and yuvām (PB 21.1.1).
3.2 We obtain further results by applying more strictly our rule that Pāņini's grammar is to be taken seriously. Grammatical sūtras that are not indicated as being optional must be accepted as intended to be of general validity. In incidental cases this may give rise to doubts, but no such doubt seems to attach to the following cases.
P.2.3.61 (preşyabruvor haviso devatāsampradāne) is a rule valid for Brāhmana literature (anuvytti of brähmane from rule 60; see Joshi and Roodbergen 1981, 101, n. 331), prescribing a genitive for the object of prepya and brü, if it is an oblation in an offering to a deity. It thus excludes the use of the accusative in such cases. Yet the accusative is often used in the Satapatha Brahmana, most clearly in agnişomäbhyam chāgasya vapäm medah presya (SB 3.8.2.27; SBK 4.8.2.21), agnişomābhyām chagasya havih preşya (SB 3.8.3.29; SBK 4.8.3.18), indrāya somān prasthitān preşya (SB 4.2.1.23; SBK 5.2.1.20), and chāgānām havih prasthitam presya (SB 5.1.3.14).22
P.3.1.59 (krmydsruhibhyas chandasi) is a nonoptional rule (cf. Kiparsky 1980, 62) prescribing an as an aorist marker after the roots kr, ms, dr, and ruh in ritual literature. It excludes in this way the forms akārşit, akārsih, akārşam, and arukșat from Vedic literature. Yet these forms occur, as follows: (a)kārşit (GB 1.3.4; CHU 6.16.1); akärsih
(SB 10.5.5.3; GB 1.3.11); akārşam (AVP 20.1.6; TB 3.7.5.5; TA 10.24.1, 10.25.1; GB 1.3.12); and arukşat (AVS 12.3.42; AVP 17.40.2).
P.4.4.105 (sabhāyāḥ yah) prescribes the suffix ya after sabha in the sense tatra sādhuh (4.4.98). The next rule, P.4.4.106 (dhas chandasi), makes an exception for ritual literature. The form sabhya derived by P.4.4.105 should apparently not occur in Vedic literature. It does, though, at the following places: AVS 8.10.9, 19.55.5; AVP 16.133.5; MS 1.6.11; TB 1.2.1.26, 3.7.4.6; and SB 12.9.2.3.
P.5.4.103 (anasantān napumsakāc chandasi) prescribes for ritual literature the addition of tac to neuter Tatpuruşa compounds the last member of which end in an or-23. Patañjali in his Mahābhāşya (2:441) makes this rule optional, in order to account for words like brahmasāman and devacchandas, but this merely emphasizes the fact that Pāņini's rule is not optional. Yet there are numerous exceptions, some of which occur in the following texts:
AVS 5.10.1-7 (aśmavarman), 19.7.2 (mygasiras), 19.30.3 (devavarman).
AVP 5.29.1 (suryavarcas), 6.12.9-11 and 6.13.1-3 (aśmavarman), 13.11.21 (devavarman), 19.48.14 Chiranyanāman).
MS 3.6.7 (dikṣitavāsas), 3.11.9 (vyāghraloman). VSM 19.92 (vyāghraloman - MS 3.11.9). VSK 21.6.13 (väghraloman - MS 3.11.9 and VSM 19.92).
AB 1.26 (devavarman), 4.19 (brahmasāman, agnistomasäman), 7.19 (işudhanvan), 8.5 and 8.6 (vyāghracarman).
KB 2.1, 5.7, and 27.1 (devakarman), 5.5 (pürvedyuḥkarman and pustikarman), 5.7 (pitȚkarman), 8.7 (pasukarman), 27.1 (agniştomasāman), 30.11 (rātricchandas).

Page #11

GB 1.3.16 (sarvacchandas), 1.5.25 (svakarman), 2.1.23 (pustikarman, pūrvedyuhkarman), 2.6.6 (yajña parvan).
TB 1.7.8.1 (śärdülacarman).
SB 4.6.6.5 and 13.3.3.5 (brahmasaman), 5.3.5.3, 5.4.1.9, and 11 (sārdülacarman), 6.6.1.4, 7.3.1.4, etc. (adhvarakar. man, agnikarman), 13.3.3.4 (maiträvarunasäman), 13.3.3.6 (acchāvākasāman), 13.5.1.1 and 13.5.3.10 (agnistomasāman), 14.3.1.35 (patnikarman).
SBK 1.1.2.5-6 (mrgasiras), 7.2.4.3 and 7.3.1.9-10 (sardūlacarman).
JB 1.149, etc. (rathantarasāman), 1.155, etc. (acchävä. kasaman), 1.172, etc. (agnistomasäman), 2.240 (uttaravayas), 2.276 (ācāryakarman), etc.
PB 4.2.19, etc. (agniştomasāman), 4.3.1, etc. (brahmasaman), 8.10.1, etc. (acchāvākasāman), 9.2.7 and 15 (kşatrasāman), 9.2.20, etc. (rätrişāman), 11.3.8 and 9 (somasäman), 13.9.22 and 23 (varunasāman).
ŞadB 4.2.12-14 (brahmasaman). ĀrşB 1.378 (varunasāman), 2.3.11 (arkaširas), etc.
JĀB 5.3, etc. (somasāman), Arkaparvan 3.9 (arkasiras), etc.
SaB 1.5.15 (svakarman), 2.1.6 (setuşāman), 2.3.3 (sarpasāman), 2.3.6 (arkaširas).
SatyB, p. 72 (brahmasāman, acchāvākasāman). VamsaB 1 (girisarman). Sana 1.5 (devacchandas), 3.5 (brahmayasas, brahmatejas). TA 1.15.1, etc. (svatejas).
P.5.4.142 (chandasi ca) prescribes substitution of datæ for danta final in a Bahuvrihi compound in ritual literature. It excludes from the Vedic language Bahuvrihi com
pounds ending in danta. Yet there are some: krenadanta at AA 3.2.4 and Sana 11.4; vişadanta at AVP 5.9.8; işikadanta at AVP 1.44.2; ubhayatodanta at AA 2.3.1, SB 1.6.3.30, SBK 2.6.1.21, JB 1.128, 2.84, and 2.114, and SaB 1.8.2; and anyatodanta at SBK 2.6.1.21 and JB 1.128, 2.84, and 2.114.
P.7.1.56 (srigrāmanyos chandasi) determines the form of the genitive plural of fri and grämani as frinām and grämaninām, respectively. But genitive sūtagramanyam occurs at SB 13.4.2.5 and 13.5.2.7.
P.6.4.141 reads mantreşv any äder ātmanah (lopaḥ 134) "In mantras there is elision of the initial (sound a) of atman when (the instrumental singular ending) an follows." It is not easy to determine the precise meaning of this sütra. It may not imply that ātman never loses its initial a before other case endings, since for all we know Panini may have looked upon tman as a separate vocable, but this sütra clearly excludes the occurrence of atmana in mantras. This form is found, however, in mantras at the following places: AVS 3.29.8; AVS 5.29.6-9 - AVP 13.9.7-8; AVS 8.2.8 - AVP 16.3.9; AVS 9.5.31-36 AVP 16.99.8; AVS 18.2.7; AVS 19.33.5 12.5.5; AVP 3.28.1, 16.100.5-11, and 16.119.1-3; VSM 32.11 - VSK 35.3.8; and MS 2.8.14.
To the above cases the following may be added:
P.2.4.48 (hemantasisirāv ahorătre ca chandasi) implies, as Thieme (1935, 13) rightly pointed out, that Pāṇini "must have known fisira as a neuter." However, sisira is masculine at SVK 3.4.2; SVJ 2.3.3; AVS 6.55.2 and 12.1.36; AVP 17.4.6 and 19.9.3; SB 2.1.3.1, 2.6.1.2, 8.7.1.7 and 8, 13.6.1.10 and 11; SBK 1.1.3.1 and 1.2.3.6; JB 1.313, 2.51, 2.211, 2.356; and TA 1.6.1.

Page #12

P.3.1.118 (pratyapibhyam graheḥ [without chandasi; see Kielhorn 1885, 192 (195); Thieme 1935, 16]) prescribes pratigrhya- and apigṛhya-. Katyāyana's varttika on this sutra confines it to Vedic literature (chandas) and Patanjali mentions the alternatives pratigrāhya- and apigrahya-. The last two forms were apparently not known to Pāņini, yet apratigrahya- occurs at SaB 1.7.2.
4. What patterns arise from these data? Which Vedic texts did Panini know, and which did he not know? We shall try to arrive at an opinion on the basis of the forms
emphatically accepted or rejected by Panini himself.24
4.1 Panini records a number of forms that occur in the Ṛgveda and nowhere else. Among the forms he clearly rejects, not one occurs in the Ṛgveda. To this must be added the fact that P.1.1.16-18 refer to Sakalya's Padapāṭha. The Padapatha was added to the collection of hymns (excepting six verses; see Kashikar 1951, 44) and presupposes the latter. We may safely assume that Pāņini knew the collected Ṛgveda, not just the individual hymns.
Note that this is in no way obvious. Pāņini knew Vedic stanzas (rc) and sacrificial formulas in prose (yajus)--both of these went by the term mantra--besides brāhmaṇa and kalpa. He nowhere says that he knew the mantras in collections. In this connection it is interesting to observe that the term that came to designate such collections (samhita) did not yet have this meaning in Panini's grammar and in the Vedic scriptures. There it is synonymous throughout with sandhi. The samhita-patha, as opposed to the pada-patha, is the version of the text with sandhi.
4.2 The question as to whether the Vedic collections,
96
the Samhitas, existed in Panini's time as collections becomes pertinent with regard to the Taittiriya Samhita. We saw that three forms prescribed by Panini occur in the Taittiriya Samhita. and nowhere else (2.2, above). All these words occur in mantras. This means that possibly Panini may not have known the brahmana portions of the Taittiriya Samhita. This possibility is supported by the fact that these brahmana parts frequently contain a conspicuous non-Papinian feature, viz., the ending -ai instead of -as (see Caland 1927, 50; Keith 1914, 1:cxlv f.). Note also that the brahmana portion of the Taittiriya Samhita refers twice (6.1.9.2, 6.4.5.1) to Aruna Aupavesi, whose grandson Svetaketu Aruneya is characterized as modern in the Apastamba Dharma Sutra (1.5.5).
All this suggests that the Taittiriya Samhita was collected in its more or less final form at a late date, perhaps later than Panini. This agrees with some facts regarding the Taittiriya Brahmana and Taittiriya Aranyaka, to which we now turn.
Both the Taittiriya Brahmana and the Taittiriya Aranyaka contain forms that are explicitly rejected by Panini. The Taittiriya Brahmana has idävatsarīna, anuvatsarina, itarad (3.1, above), akārşam, sabhya; and sārdulacarman (3.2). The Taittiriya Aranyaka has akārṣam, svatejas, and sisira (m.) (3.2). It seems safe to conclude that these works were not known to, or accepted by, Panini. The Baudhayana and Apastamba Srauta Sutras "accord in recognizing the whole content both of the Brahmana and of the Aranyaka" (Keith 1914, 1:lxxviii). Yet "it would be impossible, so far as can be seen, to prove that to [these Sutras] even the
97

Page #13

Sanhită was yet a definite unit" (ibid., lucix-lxxx). The sütras only distinguish between mantra and brähmana, which occur in each of the three, Taittiriya Samhitā, Taittiriya Brāhmaṇa, and Taittiriya Aranyaka.25
The interrelationship of mantras and brāhmana portions of the three Taittiriya texts suggests that they, or parts of them, once existed as an undivided whole. We see, for example, that the brāhmana portions of TS 2.5.7 and 8 comment on the mantras of TB 3.5.1 and 2: TS 2.5.9 on TB 3.5.3.1-4.1; TS 2.6.1 and 2 on TB 3.5.5-7; TS 2.6.7 on TB 3.5.8; TS 2.6.9 on TB 3.5.10; and TS 2.6.10 on TB 3.5.11 (Keith 1914, 1:lxxxiv). TS 3.5.11 supplements TB 3.6.1, giving the mantras for the hot for the animal sacrifice (Keith 1914, 1:286, n. 4). Keith (1914, 1:1xxix) comes to a similar conclusion on the basis of the Srauta Sutras: "So far as we can judge there is no trace of any distinction being felt by the Sūtrakāras between the nature of the texts before them."
It is not impossible that the creation of a Pada pätha differentiated the Taittiriya Samhitā from Taittiriya Brähmana and Taittiriya Aranyaka, just as the Rgveda may conceivably have been collected by the author of its Padapātha (Bronkhorst 1982a, 187).
The fact that Pāṇini derives the term Taittiriya, in the sense 'uttered by Tittiri', in P.4.3.102 does not, of course, prove that the Taittiriya texts, as now known, were known to him. Panini probably knew the mantras, or a number of them, that are now part of the Taittiriya Samhită, and he may indeed have considered them taittiriya 'uttered by Tittiri'.
Note, finally, that the Taittiriya Samhitā appears to borrow from the Aitareya Brāhmana 1-5, as argued by Keith (1914, 1:xcvii f.); see also Aufrecht (1879, vi, 431f.) and Keith (1920, 46). The Aitareya Brāhmana itself, including its first five chapters, deviates in a number of points from Pāņini (4.5, below).
4.3 Some of the other Samhitās of the Yajurveda sin occasionally against Păņini.
The Vājasaneyi Samhitā has ātmanā, masculine sisira, and one Tatpuruşa compound in -an (vyāghraloman). It shares this, however, with the Maitrāyani Samhita.
The Maitrāyani Samhita has sabhya, some Tatpuruşa compounds in -as and -an, ātmanā; note further dādhrati (see note 19, above). These deviations from Pānini in the Maitrāyani Samhitā are most surprising because Panini appeared to know both the mantra and brāhmana portions of this text (see 2.3, above). This warns us once again that we cannot assume that the texts we know now existed in the same form in Pāņini's day.
4.4 Did Pānini know the Atharvaveda? Two forms prescribed by him are found only there, one in the Saunakiya version and one in the Paippalāda version. However, opposed to these two forms are numerous others forbidden by Pāṇini. They include gamayām cakāra, gamayan cakartha (3.1), akārsam, arukşat, sabhya, several neuter Tatpuruşa compounds ending in -an and -23, vişadanta and işikādanta, haricandra, ātmanā, and sisira (masc.) (3.2).
One might raise the question of whether the word-forms in the Atharvaveda may not have been Vedic in Pāņini's opinion, that is, whether, perhaps, they were covered by

Page #14

non-Vedic rules of the Astādhyāyi. This is suggested by Balasubrahmanyam's remark (1984, 23):
Among the seven khyun- derivatives taught by Plānini) in Alstādhyāyi) 3.2.56, subhagamkarani and priyam-karanam are only attested in the Samhita texts of the (Atharvaveda)--the former occurring at (AVS) 6.139.1 and AVP 7.12.5, and the latter at the Paippalada Samhita (3.28.5; 6). Neither in the other Vedic Samhitās nor in the Brahmana-Aranyaka texts, do we come across these derivatives.
Balasubrahmanyam's observation is misleading in that subhagamkarani is not taught in P.3.2.56 nor anywhere else in the Astädhyāyi. This is so because a vārttika of the Saunāgas (Mbh, 2:105, 1.8; on P.3.2.56) is required to provide subhagamkarana with its feminine ending, i, as shown by Balasubrahmanyam himself. Thus, P.3.2.56 did not derive subhagamkarani in the Atharvaveda. The fact that the Atharvaveda contains two more words of the same kind (ayakşmamkarani at AVS 19.2.5 and AVP 8.8.11; sarūpamkarani at AVS 1.24.4 and AVP 1.26.5; see Balasubrahmanyam 1984, 25f.) and that these words are not even partiallyco derived in Pāņini's grammar, makes it less than likely that the priyamkaranam of AVP 3.28.6 was meant to be explained in P.3.2.56.
An interesting confirmation that the Atharvaveda did not exist as a collection until long after the other three Vedas were collected is found in the Chandogya Upanişad. Sections 3.1-5 make a number of comparisons, or rather identifications, of which the following are of interest to us.
Section 3.1 states that the bees are the res, the flower is the Rgueda; in 3.2 the bees are the yajus (pl.), the flower is the Yajurveda; and in 3.3 the bees are the sämans, the flower is the Sāmaveda. The interesting observation comes in section 3.4, where the bees are the atharvangirasah and the flower is itihäsa purānam. In 3.5, finally, the bees are the hidden teachings (guhyā ādesah), which may be the Upanişads, and the flower is Brahman (n.). Since the athar. vāngirasah constitute the Atharvaveda as we know it, the logic of the situation would have required that the flower in 3.4 be identified with the Atharvaveda. The fact that it is not hardly allows an explanation other than that the author of this passage did not know of such a definite collection of atharvans and angirases. Itihāsa and purāna certainly do not designate the Atharvaveda, neither separately nor jointly (see Horsch 1966, 13f.).
Bloomfield (1899, 2f.), too, came to the conclusion that many hymns and prose pieces in the AV. date from a very late period of Vedic productivity." Indeed, there is nothing in the way of assuming that the composition of such texts as the AB. and SB. preceded the redactions of the Atharvan Samphitās."
Patañjali's Mahābhāşya cites in its opening passage the first lines of the four Vedas; these apparently existed as collections in those days (second century B.C.). The first line is sam no devir abhiştaye, which begins the Paippalada version of the Atharvaveda. Patañjali even informs us of the size of the Atharvaveda known to him, saying (Mbh, 2:378, 1.11; on P.5.2.37): vimsino'ngirasah. This fits the twenty books of the Paippalāda Samhita. We may conclude
100
101

Page #15

from this that the Paippalāda Samhitä existed in its present form, at any rate, in the second century B.C.
4.5 The Aitareya Brāhmana transgresses Pāņini's rules in containing itarad, nominative avām (3.1), and several neuter Tatpuruşa compounds in -an (3.2). It is also interesting that AB 7.17 has the periphrastic perfect amantrayam äsa, as opposed to P.3.1.40, which allows only ky in such formations (Keith 1936, 747). We also find optatives in (aylita instead of (ay)eta (Renou 1940, 11), and the ending -ai for both genitive and ablative -as (Caland 1927, 50), not prescribed by Pāņini.
By way of exception some older arguments adduced by Keith (1920, 42.) to determine the date of the Aitareya Brāhmana will be reviewed here (see also Bronkhorst 1982, 276). The language of this Brāhmana is said to be "decidedly older than the Bhäşā of Pāņini," on the basis of Liebich's Pānini (1891). The circularity of Liebich's arguments has been shown elsewhere (Bronkhorst 1982, 275f.). The fact that Yāska knew the Aitareya Brähmana is irrelevant, since it is very likely that he is later than Panini (Bronkhorst 1984, 8f.). The Aitareya Brāhmaṇa contains indications that it knew the Rgveda before the completion of the orthoepic diaskeuasis but this implies nothing in view of the fact that the orthoepic diaskeuasis of the Rgveda was not completed until long after Pāṇini (Bronkhorst 1981). The absence of reference to metempsychosis must be viewed against the background of the unwillingness of orthodox Brahmanism to let these ideas find entrance into their sacred texts even at a time when they had become generally known and widely accepted (Bronkhorst 1989, 125).
4.6 The other Brahmanas that are considered early are the Kausitaki Brāhmana, Pancavimsa Brähmana, Jaiminiya Brähmana, and Sata patha Brāhmana (Renou 1957, 14). We can be brief about them..
The Kausitaki Brähmana has a number of forbidden words: saprabhyti, sodarka, and itarad, besides many neuter Tatpuruşa compounds in -an and at least one in -as. Like the Aitareya Brāhmana, it has optatives in (ay)ita and -ai for -ās.
The Pañcavimsa Brahmana, too, has saprabhyti and sodarka, as well as nominative yuvām, and many neuter Tatpuruşa compounds in -an.
The Jaiminiya Brähmana goes against Panini's grammar in having itarad, many neuter Tatpuruşa compounds in -an and -as, ubhayatodanta and anyatodanta, and masculine sisira.
The Satapatha Brāhmana deviates from Panini's grammar in the words itarad, nominative avām, akārsih, sabhya, an accusative rather than a genitive for the object of preşya, many neuter Tatpuruşa compounds in -an, ubhayatodanta, genitive plural -grāmanyam, and masculine sisira.
The Kānva version of the Sata patha Brāhmana, finally, deviates in fewer respects, containing a few neuter Tatpuruşa compounds in -an and -as, ubhayatodanta and anyatodanta, an accusative rather than a genitive for the object of preşya, and masculine sifira.
5. The above considerations must be treated with caution. For one thing, it is not known in any detail what changes were made in the texts during the process we refer to as their worthoepic diaskeuasis." This implies that we cannot be altogether sure what features of those texts can
102
103

Page #16

nowhere else, but this proves no more than that Panini was acquainted with certain portions of them, if it proves any. thing at all.
The regional origin and early spread of most of the Vedic texts may account for Pānini's lack of acquaintance with some of them. Panini is held to have lived in northwest India. Texts from other parts of the country may only have become known to him if they were generally accepted as Vedic in their region and beyond it.
Notes
be used to determine their relationship with Panini's Astädhyayi. We also do not know how many serious deviations from Panini's explicit statements must be considered evidence that Pānini was ignorant of a particular text. Further, we should remember that we made an assumption, which may not be acceptable to everyone, that Panini's, grammar can be taken seriously.
Neither should we be rash in concluding that Vedic texts that transgress the rules of Pāņini repeatedly were for that reason completely unknown to Pāņini. The problem is that, probably, no Vedic text has a single author. All are collections of parts of more or less heterogeneous origin. This applies to the Samhitās as well as to the Brähmanas and Aranyakas. The most we can conclude from the deviations between the majority of Vedic texts and Panini's grammar is that Päņini did not know much of Vedic literature in its present form, that is, in the collections known to us. Much of Vedic literature was still in a state of flux in Panini's day, and had not yet reached the unalterable shape in which we know it.
These considerations are of value with regard to the texts that would seem to have been unknown to Panini on the basis of the evidence reviewed in this article. They are, however, of equal value where the texts that appear to have been known to Pāṇini are concerned. The Rgveda may be an exception; it was known to Pāņini along with its Padapatha, which leaves little room for major changes other than sandhi. But we must be cautious with respect to such texts as the Maitrāyani Samhita and Kathaka Samhitā. It is true that they contain words prescribed by Panini, which occur
This study was carried out as part of a project of professors M. Witzel and T. E. Vetter, which was financed by the Netherlands Organization for the Advancement of Pure Research (Z.W.O.). In particular, Professor Witzel took a lively interest in the project. One of his own fields of specialization is the geographical distribution of Vedic schools in different periods. It is hoped that from that side additional evidence will come forth to shed light on the problems discussed here. Meanwhile, Witzel's "Tracing the Vedic Dialects" (1989) has appeared, which, unfortunately, could not be taken into
consideration for the present article. 1. For a survey, see Cardona (1976, 226-28). Some important
articles have been reproduced and discussed in Staal
(1972, 135-204) 2. This means that one cannot conclude from certain peculi
arities of sandhi in the Maitrāyani Samhitā which are
104
105

Page #17

not described by Panini, that they "escaped his observa
tion," as Palsule (1982, 188) claims. 3. Balasubrahmanyam (1981, 400) notes that in the sample
studied by him, "three per cent of the exclusive Vedic vocabulary differs from Plāņini)'s accentual system, and four per cent of the common vocabulary manifests the apparent difference between P[anini) and the Veda with
reference to the systems of Kyt accentuation." 4. Even Kätyāyana and Patañjali sometimes ascribe an accent
to a Vedic word that deviates from the accent found in the surviving texts (see Balasubrahmanyam 1974, 3, on
sthāsnu). 5. The fact that the Phitsūtras of Santanava ascribe to
arya, jyestha, and kaniştha the accents found in the extant Vedic literature is reason to think that Säntanava is later rather than earlier than Panini; cf. Kielhorn (1866, 1f.) and Devasthali (1967, 39f.). Kapila Deva Shastri (Sam 2018, 28f.) argues for an earlier date of the Phitsutras on insufficient grounds (Cardona 1976,
176). 6. Cf. Kiparsky (1980, 69) and Devasthali (1984, 137). 7. Thieme (1985) shows that the accents prescribed by
Panini in the case of words that are commonly used to address people are the initial accents of the vocative. He concludes that Panini's accents are later than the (differing) Vedic ones. This may be correct, yet it does not by itself prove that all the texts having Vedic accentuation in these cases are older than Panini. It is certainly conceivable that the Vedic texts were composed in a form of language that was kept archaic also
in its accents. Pāņini's bhāṣā, too, is younger than Vedic, yet Panini does not for that reason necessarily
postdate scriptures that use the Vedic language. 8. Note that MS-1.5.12 (p. 81 1.2-6) uses rätri in the
language used by the gods and rätri elsewhere; this was
pointed out to me by Professor Witzel. 9. P.2.3.62, 4.39, 73, 76, 3.2.88, 5.2.122, 6.1.34, 70, 133,
178, 2.199, 4.75, 7.1.8, 10, 103, 3.97, 4.78. Cf. Shiva
ramaiah (1969) 10. It goes without saying that the generality of such rules
can be restricted in various ways such as the presence of
rules that account for exceptions (apavāda). 11. For a brief description, see Chakrabarti (1980, 134-36)
and Jha (1942, 294-99). 12. The following examples are taken from Bharthari's dis
cussion of iha in his commentary on the Mahābhāşya (see
below). 13. We should not be misguided by this late date. Bhart
hari made use of works on Mimamsă older than Sabara's, among them probably the one by Bhavadāsa. See Bronk
horst (1989a). 14. The relevance for grammar is, of course, that in this way
it can be decided whether or not Vedic rules are to be used in the modified mantras. Note that Kumărila's Tantravārttika on PMS 1.3.24 maintains that üha is brought about without the help of grammar but rather with
forms found in the Veda. 15. SSS 6.1.15, similarly, prescribes substitution of medha
pataye or medhapatibhyaḥ for, apparently, medha patibhyām, as instances of wha.
106
107.

Page #18

16. ghas replaces ad before aorist endings according to
P.2.4.37 (lunsanor ghas!). 17. The value of this case is somewhat in doubt since TB
2.8.2.5 cites the same mantra as RV 10.47.1 with
jagrohna; it may have contained jagrbhma. 18. My friend Dr. Harry Falk points out in a forthcoming
article that Panini cannot have known AV$ 6.16.3, the reason being that Pāņini derives ailayit from elayati: the same mantra contains the form ilaya (avelaya), so that Panini, had he known AVS 6.16.3 as a whole, would have derived ailayit from ilayati rather than from
elayati. 19. This was pointed out by Manjul Mayank in a paper read at
the Seventh World Sanskrit Conference, Leiden, 1987. 20. The corresponding plural dădhrati occurs at TS 2.3.1.2,
5.3.9.2; MS 2.2.1; and KS 11.6. However, the juxtaposi. tion of dadharti, dardharti, dardharşi, and other finite verb forms seems to indicate that the precise form
dādharti is meant. 21. For example, P.7.1.57 goh pädänte) prescribes that the
genitive plural of go at the end of a verse-foot in ritual literature is gonām. This is illustrated in RV 10.47.1. But the Käsikā rightly observes that there are exceptions: RV 10.166.1 has gavām at the end of a verse
foot. 22. The Kānva parallel SBK 6.1.3.12 (chägänäm havişām
prasthitam preşya) seems to be the only example in Vedic literature in which P.2.3.61 is obeyed. Note that the single vārttika on P.2.3.61 is intended to make the rule invalid where the oblation is prasthita. This would
justify all, or almost all, deviations from Panini's rule, yet the fact that Panini says nothing about prasthita in this context shows that he did not know, or accept, these counterexamples. Similarly, see Navathe
(1987). 23. bähvojas in RV 8.93.2 is considered a Bahuvrihi, and not
therefore a Tatpuruşa compound, by Oldenberg (1909-12, 2:144). soma parvabhih in RV 1.9.1 - AVS 20.71.7 - VSM 33.25 - VSK 32.2.8 - SVK 1.180 = SVJ 1.2.1.7.6 can be
derived from parva, by P.7.1.10. 24. Note that the insufficiency of Păņini's grammar with
regard to the Vedic data has been known for a long time in the Pāṇinian tradition. Kumārila Bhatta, in his Tantravärttika, cites in this connection SVK 2.1006 = SVJ 4.17.11 (madhya apasya tisthati), which has apasya
instead of apām. 25. Caland (1921, 3) observed that the Āpastamba Srauta Sutra
refers to mantras of the Taittiriya Samhita by way of their initial words, and to those of the Taittiriya Brähmana by citing them in full. Kashikar (1968, 400) has also shown that mantras from the Taittiriya Brāhmana are often quoted by pratika. The Bharadvä ja Srauta Sutra
follows a similar practice (Kashikar 1968, 401). 26. That is, not even the forms ayakşmamkarana and sarūpam
karana, without the feminine i, are derived. 27. Note that the Mahābhāşya also prefers the Paippalāda
version of the Atharvaveda in some citations (see Renou 1953, 463).
108
109

Page #19

Abbreviations
SaB Sāmavidhāna Brāhmana ŞadB Şadvimsa Brāhmana, edited by Bellikoth Ramachandra
Sharma. Tirupati: Kendriya Sanskrit Vidyapeetha.
1967. SãGS Sankhāyana Grhya Sūtra Sana Sankhāyana Aranyaka SatyB sāļyāyana Brāhmana SB Sata patha Brāhmana SBK Satapatha Brāhmana (Kānva) SSS Sankhāyana Srauta Sutra Sw
Swaminathan's edition of Bharthari's
Mahābhāşyadipika тв Taittiriya Brāhmana VamsaB Vamsa Brāhmana Vass Vāräha Srauta Sutra
Aitareya Aranyaka Aitareya Brāhmana Abhyankar and Limaye's edition of Bharthari's
Mahabhāşyadipika ApSS Āpastamba Srauta Sūtra Ar$B Ārşeya Brāhmana Asss Asvalāyana Srauta Sutra AVP Atharvaveda (Paippalāda) AVS Atharvaveda (Saunakiya) BAU(K) Brhadaranyaka Upanişad (Känva) Bhass Bhäradvāja Srauta Sutra CE "Critical edition" of Bharthari's Mahābhäşyadipikā Chu Chandogya Upanişad DrŚs Drāhyāyana Srauta Sūtra HISS Hiranyake i Srauta Sutra JAB Jaiminiya-Ārseya-Brāhmana, edited by Bellikoth
Ramachandra Sharma. Tirupati: Kendriya Sanskrit
Vidyapeetha. 1967 JUB Jaiminiya Upanişad Brāhmana KB Kauşitaki Brähmana
Kathaka Samhita LSS Latyayana Srauta Sutra Ms Manuscript of Bharthari's Mahābhāşyadipika MS Maitrāyani Samhita MÁS Mänava Srauta Sutra
Pāninian sutra
Pancavimsa Brāhmana PMS Purva Mimāmsā Sutra
Bibliography
Aufrecht, Theodor, 1879: Das Aitareya Brāhmana. Bonn:
Adolph Marcus.
KS
Balasubrahmanyam, M. D., 1964: "The accentuation of arya in
Panini and the Veda." 26th International Congress of Orientalists. New Delhi. Summaries of Papers, III, 54-55.
P.
PB
..., 1966: "An accentual problem in Panini and the Veda
apropos of the word hāyana-." Bulletin of the Deccan College Research Institute 25, 43-58.
110
111

Page #20

----
1969: "Arya-: an accentual study." Indian Antiquary, Third Series, 3 (R. N. Dandekar Felicitation Volume), 112-27.
1972: "Vedic śriyase and Panini 3.4.9." Vishveshvarananda Indological Journal 10, 7-10.
1974: "Panini 6.1.209-210." Charudeva Shastri Felicitation Volume. Delhi, 189-93.
- 1974a: "Kätyāyana and some Vedic formations." Samskṛtavimarśaḥ 1(2), 1-4.
, 1981: The System of Krt Accentuation in Panini and the Veda. Tirupati: Kendriya Sanskrit Vidyapeetha (Kendriya Sanskrit Vidyapeetha Series No. 32).
1983: "An accentual note on Vedic voḍhave." Surabhi. Sreekrishna Sarma Felicitation Volume. Tirupati: Sri Venkateswara University, 229-40.
" 1984: "Subhagamkarani in AV 6,139,1." Amṛtadhārā. R. N. Dandekar Felicitation Volume. Delhi: Ajanta Publications, 21-27.
Bhartrhari: Mahabhäşya Dipikā. (1) Edited by K. V. Abhyankar and V. P. Limaye. Poona: Bhandarkar Oriental Research Institute. 1970 (Post-Graduate and Research Department Series No. 8). (2) Partly edited by V. Swaminathan under the title Mahabhāṣya Tikā. Varanasi:
112
Banaras Hindu University. 1965 (Hindu Vishvavidyalaya Nepal Rajya Sanskrit Series Vol. 11). (3) Manuscript reproduced. Poona: Bhandarkar Oriental Research Institute. 1980. (4) "Critical edition" and translation. Poona: Bhandarkar Oriental Research Institute. To date, six volumes have been published: Ahnika 3 by G. B. Palsule (1983); Ahnika 5 by V. P. Limaye, G. B. Palsule, and V. B. Bhagavat (1984); Ahnika 6, part 1, by V. B. Bhagavat and Saroja Bhate (1986); Ahnika 1 by Johannes Bronkhorst (1987); Ahnika 2 by G. B. Palsule (1988); and Ahnika 4 by G. V. Devasthali and G. B. Palsule (1989).
Bhat, M. S., 1968: "The Vedic stem ratri- and Panini." Journal of the Bombay Branch of the Royal Asiatic Society 41-42 (1966-67, N.S.), 8-11.
Bloomfield, Maurice, 1899: The Atharva Veda. Asian Publication Services. 1978.
1906: A Vedic Concordance. Delhi, Varanasi, and Patna: Motilal Banarsidass. 1964 (Harvard Oriental Series, Vol. 10).
Bronkhorst, Johannes, 1981: "The orthoepic diaskeuasis of the Rgveda and the date of Panini." Indo-Iranian Journal 23, 83-95.
1982: "The variationist Panini and Vedic." IndoIranian Journal 24, 273-82.
113

Page #21

......, 1982a: "Some observations on the Padapātha of the
Rgveda." Indo-Iranian Journal 24, 181-89.
Chakrabarti, Samiran Chandra, 1980: The Paribhāşās in the
Srautasütras. Calcutta: Sanskrit Pustak Bhandar.
-----, 1984: "Nirukta, Uņādi Satra, and Aştādhyāyī."
Indo-Iranian Journal 27, 1-15.
Devasthali, G. V., 1967: Phițsūtras of Santanava. Edited with
introduction, translation, and critical and exegetical notes. Poona: University of Poona. (Publications of the Centre of Advanced Study in Sanskrit, Class C, No. 1).
......., 1989: "L'indianisme et les préjugés occidentaux."
Etudes de Lettres (Lausanne), April-June 1989, 119-36.
---, 1989a: "Bharthari and Mimāmsā." Studien zur Indo
logie und Iranistik 15, 101-17.
......., 1984: "Panini and Vedic-a critique." Annals of
the Bhandarkar Oriental Research Institute 64 (1983), 137-48.
Caland, W. (tr.), 1921: Das Srautasūtra des Āpastamba.
Göttingen: Vandenhoeck and Ruprecht.
Gonda, Jan, 1975: Vedic Literature (Samhitās and Brāhmanas).
Wiesbaden: Otto Harrassowitz (A History of Indian Literature, Vol. I, Fasc. 1).
........ 1927: "On a paragraph of Vaidic syntax." Acta
Orientalia 5, 49-51.
Goto, Toshifumi, 1987: Die "1. Präsensklasse" im Vedischen.
Wien: Verlag der Österreichischen Akademie der Wissenschaften.
..., 1928: “Eine vierte Mitteilung über das Vādhala
sūtra." Acta Orientalia 6, 97-241.
Hoffman, Karl, 1974: "Panini VII 2, 69 sanim sasanivāmsam."
Münchener Studien zur Sprachwissenschaft 32, 73-80.
(tr.), 1928a: Das Srautasütra des Āpastamba. Sechszehntes bis vierundzwanzigstes und einunddreissigstes Buch. Amsterdam: Koninklijke Akademie van Wetenschappen.
Horsch, Paul, 1966: Die vedische Gatha- und Sloka-Literatur.
Bern: Francke Verlag
Cardona, George, 1976: Panini: A Survey of Research. Delhi,
Varanasi, and Patna: Motilal Banarsidass. 1980.
Jha, Ganganatha, 1942: Pürva-Mimāmsā in its Sources. Second
edition, 1964. Varanasi: Banaras Hindu University.
114
115

Page #22

1936: “Pāṇini and the Veda.” Indian Culture 2, 735-48.
Joshi, S. D., and Roodbergen, J. A. F., 1981: Patanjali's
Vyakarana Mahābhāşya, Prätipadikärthasepähnika (P.2.3.46-2.3.71). Pune: University of Poona (Publications of the Centre of Advanced Study in Sanskrit, Class C, No. 14).
Kielhorn, Franz, 1866: Çāntanava's Phitsūtra. Mit verschie
denen indischen Commentaren, Einleitung, Uebersetzung und Anmerkungen herausgegeben. Genehmigter Nachdruck. Nendeln, Liechtenstein: Kraus Reprint, Ltd. 1966 (Abhandlungen der Deutschen Morgenländischen Gesellschaft, IV. Band No. 2).
Kapila Deva Shastri, Sam 2018: Samskrta Vyakarana mem
Gana pātha ki Paramparā aura Ācārya Pāṇini. Ajmer: Bhāratiya-Prācyavidya-Pratişthāna.
Kashikar, C. G., 1951: "The problem of the Galantas in the
Rgvedapadapātha." Proceedings of the All-India Oriental Conference 13 (1946), 39-46.
...., 1885: "Der Grammatiker Panini." Göttinger Nachr.
1885, 185-99. Reprinted in Kleine Schriften I, 188202. Wiesbaden: Franz Steiner, 1969.
1968: "The Taittiriya-Brähmana in relation to the Satrakāras." Pratidānama. Festschrift Franciscus Bernardus Jacobus Kuiper. The Hague and Paris: Mouton, 398-408
Kiparsky, Paul, 1980: Panini as a Variationist. Edited by
S. D. Joshi. Pune: Centre of Advanced Study in Sanskrit, in collaboration with the MIT Press, Cambridge, Mass., and London.
......., 1982: Some Theoretical Problems in Panini's Gram
mar. Poona: Bhandarkar Oriental Research Institute.
Keith, Arthur Berriedale (tr.), 1914: The Veda of the Black
Yajus School entitled Taittiriya Sanhita. 2 parts. Cambridge, Mass.: Harvard University Press (Harvard Oriental Series 18, 19).
Kumārila Bhatta: Tantravärttika. In: Mimāmsādarsana. Edited
by Kašinātha Vasudevaśāstri Abhyamkara and Pt. Ganesaśāstri Josi. Poona: Anandāśrama. 1973-84 (Anandāśrama Samskstagranthāvali 97).
--, (tr.), 1920: Rigveda Brahmanas: The Aitareya and
Kausitaki Brāhmanas of the Rigveda. Delhi, Patna, and Varanasi: Motilal Banarsidass. 1971 (Harvard Oriental Series 25).
Liebich, Bruno, 1891: Panini: Ein Beitrag zur Kenntnis der
indischen Literatur und Grammatik. Leipzig: H. Haessel.
116
117

Page #23

Mayank, Manjul, 1990: "Pāṇini's acquaintance with the
Atharvaveda." In: Panini and the Veda. Edited by Madhav M. Deshpande. Leiden: E. J. Brill.
......., 1953: “Le Veda chez Patañjali.” Journal Asiatique
241, 427-64.
Navathe, P. D., 1987: "On the prasthitam havih." Annals of
the Bhandarkar Oriental Research Institute 68 (Ramakrishna Gopal Bhandarkar 150th Birth Anniversary Volume), 645-51.
---, 1957: Altindische Grammatik: Introduction générale.
Nouvelle édition du texte (by Jacob Wackernagel) paru en 1896, au tome I. Göttingen: Vandenhoeck and Ruprecht
Scheftelowitz, Isidor, 1906: Die Apokryphen des Rgveda.
Hildesheim: Georg Olms Verlagsbuchhandlung. 1966.
Oldenberg, Hermann, 1909-1912: Rgveda. Textkritische und
exegetische Noten. 2 vols. Berlin: Weidmannsche Buchhandlung (Abhandlungen der königlichen Gesellschaft der Wissenschaften zu Göttingen, phil.-hist. Klasse, N.F. Band XI Nro. 5 and Band XIII Nro. 3).
-1919: "Die Nividas und Praişas, die ältesten
vedischen Prosatexte." Zeitschrift der Deutschen Morgenländischen Gesellschaft 73, 30-50.
Palsule, G. B., 1982: "Two Vedic rules of Panini: 8.3.47 and
8.3.10." Golden Jubilee Volume. Poona: Vaidika Samsodhana Mandala, 185-88.
Schroeder, Leopold von, 1879: "Ueber die Maitrāyani Samhita,
ihr Alter, ihr Verhältniss zu den verwandten Çākhā's, ihre sprachliche und historische Bedeutung." Zeitschrift der Deutschen Morgenländischen Gesellschaft 33, 177-207.
Paniniya Sikşā. Edited by Manomohan Ghosh. Calcutta:
University of Calcutta. 1938.
Patañjali: Vyākarana-Mahābhāşya. Edited by F. Kielhorn.
Third Edition by K. V. Abhyankar. 3 vols. Poona: Bhandarkar Oriental Research Institute. 1962-72.
1880: "Das Kathakam und die Maitrāyani Samhita." Monatsberichte der Königlich Preussischen Akademie der Wissenschaften zu Berlin (1879), 675-704.
------, (ed.), 1881-86: Maitrāyani Samhita. Parts I-IV.
Leipzig: F. A. Brockhaus.
Renou, Louis, 1940: "Sur certaines anomalies de l'optatif
sanskrit." Bulletin de la Société de Linguistique de Paris 41, 5-17.
---, 1895: "Das Kathaka, seine Handschriften, seine
Accentuation und seine Beziehung zu den indischen Lexi
118
119

Page #24
cographen und Grammatikern." Zeitschrift der Deutschen Morgenlandischen Gesellschaft 49, 145-71. Witzel, M., 1989: "Tracing the Vedic dialects." In: Dialectes dans les litteratures indo-aryennes. Edited by Colette Caillat. Paris: Institut de Civilisation Indienne (Publications de l'Institut de Civilisation Indienne, Fasc. 55), 97-265. Shivaramaiah, B. K., 1969: "A note on 'bahulam chandasi." The Mysore Orientalist 2(1), 7-11. Staal, J. F. (ed.), 1972: A Reader on the Sanskrit Grammar ians. Cambridge, Mass., and London, MIT Press. Thieme, Paul, 1935: Panini and the Veda. Studies in the early history of linguistic science in India. Allahabad: Globe Press. 1938: Der Fremdling im Rgveda. Eine Studie uber die Bedeutung der Worte ari, arya, aryaman und arya. Leipzig: Deutsche Morgenlandische Gesellschaft. Reprint: Kraus Reprint, Ltd., Nendeln, Liechtenstein. 1966. 1985: "Nennformen aus Anrede und Anruf im Sanskrit." Munchener Studien zur Sprachwissenschaft 44 (Festgabe Karl Hoffmann) I, 239-58. Wackernagel, Jakob, 1896-1930: Altindische Grammatik. Gottingen: Vandenhoeck und Ruprecht. Whitney, W. D., 1893: "The Veda in Panini." Giornale della Societa Asiatica Italiana 7, 243-54. 120 121