The Metre Of The Bhaddevatā from the Journal of the Royal Asiatic Society, 1906; pp. 1 -10.


A. B. Keith, m.r.a.s.

[Clarifying additions and titles by the present editor
have been placed in square brackets]


[p. 1] It is perhaps difficult to exaggerate the importance from the point of view of the literary history of India of the Bhaddevatā attributed to Śaunaka. That this has not hitherto received full recognition is due in part to the fact that it has been held, for example even by Dr. E. Sieg, Die Sagenstoffe des gveda, pp. 126, 127. that the Bhaddevatā is later in date than the Mahābhārata. This is, however, certainly not the case, as Professor A. A. Macdonell has shown conclusively in his edition Bhaddevatā, vol. i, p. xxix. of the former work. About 300 ślokas of the work are devoted to legends, and this must, it seems, be regarded as a conclusive proof that at the date of its composition there existed in Sanskrit an ākhyāna or itihāsa literature. Now the date of the Bhaddevatā is fixed by Professor Macdonell, Op. cit. Vol. 1, pp. xxii, xxiii. Cf. Victor Henry, Revue Critique. on grounds which appear to me unassailable, at about 400 B.C., perhaps earlier. It follows, therefore, that a Sanskrit itihāsa literature can be proved to have existed in the fifth century B.C.

[p. 2] This appears to me a most important result in view of the controversy over the date of the epics. Two competing opinions on this point are held at the present time. The one, represented by such scholars as Professor Jacobi, Professor Macdonell, Professor Rapson, Cf. the discussions in J.R.A.S., 1904, pp. 435-487. and Mr. Thomas, ascribes the epics to an early date, say the sixth to the fourth century B.C., and considers that at the time of composition they were written for and were intelligible to a comparatively wide circle of the people; the other, which counts among its supporters in various degrees M. Barth, Professors Bergaigne, Lüders, and Rhys Davids, Dr. Senart, and Dr. Grierson, considers that the epics are comparatively late work, the result of the gradual growth of the influence of the literary language of the Brāhmaṇic schools, which still show in many traces evidence of their being translations or adaptations of Pāli or Prākrit originals. The question is of course intimately connected with the kindred question of the extent to which Sanskrit was ever a spoken language. It is not, I understand, ever now held that Sanskrit - in the sense of the language which was known as a bhāṣā to Pāṇinī - was a vernacular of all the people in any part of India, but it obviously makes a great difference in the view taken of the nature of Sanskrit whether we are to regard it as a mere priestly language applied in late times to secular purposes, or are to hold that there was a time when a heroic epic was written in a language approximating to that of the Kṣatriya class, and one which could be understood without great difficulty by the mass of the people. We cannot believe, I venture to think, that the early audiences to whom the epics were recited were satisfied to listen to what they did not pretend to understand. No doubt, as Dr. Grierson J.R.A.S., 1904, p. 175. says, the Rāmāyana and Mahābhārata are nowadays recited to villagers who know nothing of Sanskrit, but that is the result (a) of the sacred character now attaching to the works as the result of centuries of fame, and (b) of [p. 3] the fact that the outlines of the story are familiar through vernacular translations and imitations. Neither of these features could be found in the primitive ākhyānas out of which the epic developed. It is really inconceivable that a man should compose works to appeal to the people - as the epics were beyond question intended to do - in a language unintelligible to them, whereas there is no difficulty in understanding how the epics soon became less and less generally understood, and yet retained their hold on the populace.

Taken in this connection the Bhaddevatā appears to me to be decisive for the early date of the Sanskrit epic poetry, and against the theory of translation from Pāli or Prākrit. If there were Sanskrit epic legends in the fifth century B.C., it is unreasonable to look for the composition of the great epics in the first or second century A.D.

Since the Bhaddevatā has the great merit of being preserved in a text which is in all probability free from serious interpolation or corruption, as is proved by the quotations in the Sarvānukramaṇī, I have thought it may be of interest to examine the metres of this early piece of quasi-epic literature. In the present state of the text of the two great epics no useful comparison of metre can be made, but it is not improbable that such a comparison may in course of time be rendered possible when critical text studies of the Mahābhārata and Rāmāyaṇa have proceeded further and some better criteria of old and new strata of text have come to light.

The following remarks are based entirely on the text as constituted by Professor Macdonell, Rājendralāla Mitra’s edition being quite useless from this as from every other point of view. I use a comma to denote the cæsura, or rather diæresis, whenever it can be determined with fair certainty. It is assumed that for the purpose of the diæresis a prefix like saṁ in saṁbhūtaḥ counts as a separate word; this could easily be proved if necessary. I have omitted the references to save space, and there are very possibly some errors in the enunciation, but the main results [p. 4] will not be affected by such errors. In any case the numbers would be altered if readings other than those adopted by Professor Macdonell were accepted. Cha has, of course, been regarded as always making position.


[Anuṣṭubh / Śloka]

The anuṣṭubh in the Bhaddevatā is essentially of a later form than the anuṣṭubh, even in the latest portions, of the gveda. This is shown especially by the fact that the second pāda of each half-verse ends always For an apparent exception see Macdonell, p. xxvi, n. 2. in ⏑−⏑⏓.


In the first pada the second half is ⏑−−⏓ in 2,002 cases out of a possible 2,382 [84%]. In the first four syllables ⏑⏑⏑− and −⏑⏑− occur twice each, ⏑⏑⏑⏑ and −⏑⏑⏑ once each only. The other possible forms are all frequent. On the other hand, the remaining 380 half-verses show a much greater variety of form than is allowed in the kāvya śloka, and it seems fair to regard the śloka here as exhibiting a transition stage to the śloka of the later literature.


Of the variant forms five half-lines have nine syllables in the first pāda, which begins with ⏑⏑ save in one case (IV, 102a). In 182 cases the first pāda ends in −−−⏓. Two forms only of the first four syllables occur frequently, viz., −−⏑−−,−−⏓ in 86 cases, and ⏑−⏑−−,−−⏓ in 73 cases. In VIII, 79a occurs ⏑−⏑−−−,−⏓, but the reading may be incorrect for yathā ca gharmaḥ saṁbhūtaḥ. The other instances are distributed as follows:—

−−−−−,−−⏓ x 4

+ 2 with caesura after the 4th syllable [−−−−,−−−⏓ x 2]

⏑−−−−,−−⏓ x 5

+ 2 with caesura after the 4th syllable [⏑−−−,−−−⏓  x 2]

−⏑−−−,−−⏓ x 2

−−−⏑−,−−⏓ x 2

−−⏑⏑−,−−⏓ x 2

−⏑−⏑,−−−⏓ x 1

⏑⏑−⏑−−,−⏓ x 1

⏑⏑−,−−−−⏓ x 1


[p. 5]

In 68 cases is found −⏑⏑⏓ as the end of the first pāda. As before only two forms occur frequently, viz., −−⏑−−,⏑⏑⏓ in 16 cases + 3 with caesura after the fourth syllable [−−⏑−,−⏑⏑⏓], and ⏑−⏑−−,⏑⏑⏓ in 34 cases + 1 with caesura after the fourth syllable [⏑−⏑−,−⏑⏑⏓].


There are ten other forms, as follows:—

−−−−−,⏑⏑⏓ x 3

⏑−−−−,⏑⏑⏓ x 1

−⏑−−,−⏑⏑⏓ x 2

−−−⏑−,⏑⏑⏓ x 1

−−⏑⏑−,⏑⏑⏓ x 1

⏑−⏑⏑−,⏑⏑⏓ x 1

⏑⏑−⏑−,⏑⏑⏓ x 1

−⏑⏑−,−⏑⏑⏓ x 1

⏑⏑−−,−⏑⏑⏓ x 2

⏑⏑⏑−−,⏑⏑⏓ x 1


In 52 cases the first pāda ends in −⏑−⏓. There are seven forms, of which four are fairly common:—

−−−−,−⏑−⏓ x 16

+ 2 with caesura after the fifth syllable [−−−−−,⏑−⏓ x 2]

+ 1 with caesura after the sixth syllable [−−−−−⏑,−⏓ x 1]

⏑−−−,−⏑−⏓ x 7

−⏑−−,−⏑−⏓ x 3

+ 1 with caesura after the sixth syllable [−⏑−−−⏑,−⏓ x 1]

−−⏑−,−⏑−⏓ x 9

+ 1 with caesura after the sixth syllable [−−⏑−−⏑,−⏓ x 1]

⏑−⏑−,−⏑−⏓ x 9

+ 1 with caesura after the sixth syllable [⏑−⏑−−⏑,−⏓ x 2]

−⏑⏑−,−⏑−⏓ x 1

⏑−−⏑,−⏑−⏓ x 1


In 43 cases the first pāda ends in ⏑⏑⏑⏓. There are seven forms distributed as follows:—

[p. 6]

−−−−⏑,⏑⏑⏓ x 9

+ 2 with caesura after the fourth syllable [−−−−,⏑⏑⏑⏓ x 2]

⏑−−−,⏑⏑⏑⏓ x 8

+ 3 with caesura after the fifth syllable [⏑−−−⏑,⏑⏑⏓ x 3]

+ 2 with caesura after the sixth syllable [⏑−−−⏑⏑,⏑⏓ x 2]

−⏑−−⏑,⏑⏑⏓ x 3

−−⏑−⏑,⏑⏑⏓ x 6

+ 2 with caesura after the fourth syllable [−−⏑−,⏑⏑⏑⏓ x 2]

−⏑−⏑,⏑⏑⏑⏓ x 1

⏑−⏑−⏑,⏑⏑⏓ x 5

⏑⏑−−,⏑⏑⏑⏓ x 1

+ 1 with caesura after the fifth syllable [⏑⏑−−⏑,⏑⏑⏓ x 1]

[Javipulā or Anuṣṭupvipulā]

In 15 cases the first pāda ends in ⏑−⏑⏓. These cases are of special interest, as the later form avoids carefully the iambic ending. There are six forms:—

−−−−⏑,−⏑⏓ x 1

−⏑−−,⏑−⏑⏓ x 2

−−⏑−⏑,−⏑⏓ x 4

+ 1 with caesura after the fourth syllable [−−⏑−,⏑−⏑⏓ x 1]

−⏑−⏑,⏑−⏑⏓ x 1

⏑−⏑−,⏑−⏑⏓ x 4

a very remarkable form

⏑−⏑⏑⏑,−⏑⏓ x 1

+ 1 with caesura after the sixth syllable [⏑−⏑⏑⏑−,⏑⏓ x 1]


In 12 cases the first pāda ends in ⏑⏑−⏓. There are eight forms, but all the occurrences are sporodic:—

−−−−,⏑⏑−⏓ x 2

⏑−−−,⏑⏑−⏓ x 1

−−⏑−,⏑⏑−⏓ x 3

⏑−⏑−,⏑⏑−⏓ x 2

−−⏑⏑⏑,⏑−⏓ x 1

⏑−−⏑,⏑⏑−⏓ x 1

⏑−⏑⏑⏑⏑,−⏓ x 1

⏑⏑⏑⏑,⏑⏑−⏓ x 1


[p. 7]

There remain three cases of endings in −−⏑⏓, viz.:—

−−⏑⏑−,−⏑⏓ x 1

⏑−⏑−,−−⏑⏓ x 1

−−⏑−−,−⏑⏓ x 1

This large variety of forms appears consistent with and to support the date assigned on other grounds to the work by Professor Macdonell. It was most probably written at a time when the śloka had not yet received its final form, and when the verses which are irregular according to the later metre were still felt to be correct. It may, of course, be argued that some of the forms are the result of the introduction of quotations from the gveda, but, even allowing this to be the case in some instances, the explanation cannot be applied in the majority of cases, and it would doubtless have been easy for the author to put them in another form, had they seemed to him unmetrical.

Consistent also with the antiquity of the verse is the fact of the separation of the pādas. Hiatus is quite freely allowed between pādas in the same half-verse. There are, according to my reckoning, about 112 cases of such hiatus. It is true that hiatus occurs also elsewhere, but these cases can nearly all be reduced to (1) Vedic quotations, e.g., te astu, I, 54a; ko adya, I, 57a; or (2) a or ā + , or i + , or u + — all special cases. Śākalya, it may be noted, is cited in Pāṇini, VI, 1, 127, as permitting the absence of sandhi in the case of i, u, and followed by a dissimilar vowel, and Śaunaka is associated with Śākalya. The absence of sandhi between a or ā and is permitted by Pāṇini, VI, i, 128, also on the authority of Śākalya, according to the Kāśikā Vtti. Other exceptions are extremely rare (e.g. I, 111a). Between pādas, however, all sorts of hiatus occur freely.

On the other hand, there are not lacking signs that the connection of the pādas was becoming closer than in the period of the Saṁhitās. The instances are of three kinds.

(1) The break at the end of the first pāda occurs in the middle of a compound, or after a prefix to a verb, e.g., prātaḥ | savanam, I, 115a, or abhi | dīyate, I, 30a. There [p. 8] are seven other instances (II, 98a, lO3a; III, 86b; IV, 82b; V, 58c, 175b; VI, 88b). (2) There is elision at the end of the first pāda; the elision is almost always of i becoming y; of which there are eleven instances (II, 127b, where the verse should probably be divided after the 'py; III, 69b, 135a; IV, 144b; V, 81b; VI, 63b, 68b; VII, 83a, 105b; VIII, 14b, 94b). There are three instances of the elision of initial a (I, 54b; IV, 139a; VI, 156a); and one instance of u becoming v (II, 115b). (3) Finally, in six cases the verse runs on irregularly: they are III, 83a, āṅgirasasyāsan; 134b, varuṇasyāryamṇaḥ. 9a, naktānakti; II, 141a, hīttham; IV, 116b, aṅgānyanaḍuhaḥ. VIII, 57a, tvantyānyāḥ.



The examples of hiatus taken together with these signs of the tendency to regard the pādas as united seem to be conclusive evidence of the transitional character of the verse. The same view follows from the treatment of the triṣṭubh. There are some 42 verses in this metre in the Bhaddevatā, and the details given below seem conclusively to show that the metre was still in an experimental stage. No one after the later metres had definitely formed themselves would have composed these curious forms, which, however find a natural explanation as transitional forms from the free triṣṭubh of the Saṁhitās, where the last four syllables are alone of importance, to the later verses, where all syllables are determined.

Only one verse occurs where all the pādas are alike in metre, the indravajrā, IV, 2. Verses with mixed pādas of indravajrā and upendravajrā occur in I, 44 (u. + i. + u. + i.), 45 (u. + u. + i. + i.); IV, 1 (u. + u. + i. + i.). In III, 154, the first, second, and fourth padas are indravajrā, the third is śālinī. Editor’s note: these forms, and the similar ones listed below, are probably triṣṭubh with the Vedic opening ⏓−−−, and not the classical metre śālinī at all; the forms which show the break −⏑− are probably simply triṣṭubh with a ragaṇa break, which occurs quite often in the early triṣṭubh. In 155 the first two pādas are śālinī, the third and the fourth irregular. In IV, 95, the first and third pādas are upendravajrā and the second indravajrā, the fourth is irregular. In 96 the last two pādas are indravajrā, the first two are irregular. The second and third pādas of V, 8, are identical (⏑−−−−⏑−−⏑−⏓), and so in V, 46, and VIII, 127 (⏑−⏑−,−⏑−−⏑−⏓), the first and fourth being irregular. In [p. 9] V, 113, the first two pādas are upendravajrā, the last two ⏑−⏑−,−⏑−−⏑−⏓. In VIII, 101, the first two are indravajrā, the last irregular. In 125 the second and fourth are ⏑−−−,⏑⏑−−⏑−⏓, the others irregular. In IV, 99, the second and third are śālinī, the rest irregular. In all, 15 verses have two or more pādas alike.

On the other hand, there are no less than 7 verses with four pādas of 11 syllables all dissimilar (III, 156; IV, 5, 6, 7; V, 114; VIII, 128, 129); and 19 verses contain pādas of differing numbers of syllables. In five cases only is there any correspondence of pādas: in IV, 10, the verse consists of 12 + 11 + 11 + 11, the last two being −−−−−⏑⏑−⏑−⏓; in IV, 4, of 11 + 11 + 12 + 11, the second and fourth being −−−⏑⏑⏑−−⏑−⏓; in V, 7, of 11 + 12 + 11 + 12, the second and fourth being ⏑−−−⏑⏑−−⏑−⏑⏓; IV, 3, consists of an indravajrā + upendravajrā + indravajrā + 12; and V, 11, consists of two śālinī pādas + 12 + 11. The other verses show different variations of pādas of 10, 11, and 12 syllables as follows:

IV, 97: 10 + 11+ 11 + 11.

V, 112: 10 + 11 + 12 + 12 (the first pāda may be read as 11 with vyūha).

V, 10: 11 + 11 + 10 + 10.

III, 126, 127, VIII, 99, 100: 11 + 12 + 11 + 11.

III, 128; V, 9: 11 + 11 + 11 + 12.

IV 8: 12 + 11 + 11 + 11.

IV, 9: 12 + 11 + 12 + 11.

III, 129: 12 + 12 + 11 + 11.

III, 130: 12 + 11 + 12 + 12.

IV, 98: 11 + 11 + 12 + 12.

There remains VIII, 130, which has 6 padas of 11 syllables, the fourth and fifth being upendravajrā, and the first and sixth indravajrā.

[p. 10]


Of the 24 jagatī padas only 12 have the characteristic jagatī ending of −⏑−⏑⏓; 10 end in −⏑−⏓; Editor’s note: i.e. are triṣṭubh pādas. 2 in ⏑⏑−− and −−⏑− respectively. On the other hand, of the triṣṭubh pādas 2 end with the jagatī ending ⏑−⏑⏓, Editor’s note: i.e. are jagatī pādas. and 1 with ⏑⏑−−. In four cases hiatus is permitted between the pādas of the half-verses, while in one case tu becomes tv.


[Comparison with Aitareya Brāhmaṇa]

Confirmation of the view here taken that the metre of the Bhaddevatā represents a genuine stage of the historical development of the śloka may be derived from an examination of the 58 half-verses in the epic narrative in adhyāya 33 of the Aitareya Brāhmaṇa, which must date about 200 or 300 years before the Bhaddevatā. In 14 cases the first pāda ends in ⏑−−⏓; in 13 in −−−⏓; making 27 cases with the long syllable in the sixth and seventh places, the characteristic of the classic śloka. Of the rest there are 8 cases of ⏑−⏑− [javipulā]; 6 of −⏑−⏓ [ravipulā]; 5 of ⏑⏑⏑⏓ [navipulā]; 5 of −⏑⏑⏓ [bhavipulā]; 4 of −−⏑⏓ [tavipulā]; and 3 of ⏑⏑−⏓ [savipulā]. In three cases the second pāda has not an iambic ending. The verse is undoubtedly of an older type than that of the Bhaddevatā but the line on which it will develop is clearly one which will naturally lead to the later metre, while its own history can be traced in the different strata of the gveda.