The Oral Style of the R̥gveda

Oral Tradition, 35 (2021):3-36

1. The Study of R̥gvedic Repetitions1 

In the second volume of his 1877 edition of the R̥gveda, Theodor Aufrecht collects about three thousand repeated verses and phrases from that text. Beginning with the paired Vālakhilya hymns 8.49-52, which he describes as “two versions of the same material . . . like two school-essays” (1877:II, vii),2 Aufrecht then lists, over twenty-five pages, 176 sets of formulaic verses and variants which “express the same ideas in a somewhat different style” (1877:II, xi). At the end of the volume he appends over 150 double-columned pages of verse-beginnings and parallel passages, along with their variants in the other Vedas (1877:II, 514-666).

From their plenitude he concludes that the R̥gveda is the remnant of a long poetic tradition, in line with the Vedic belief in “an oldest or original Veda, of which the present ones are just relics.” He considers that “only few hymns are still in the form in which they were originally composed; . . . only a remnant of the ancient hymns of India survives” (1877:II, xii). This attitude is crystallized in his calling a r̥ṣi (a Vedic singer or oral poet) an epigone (1877:II, xxiv).

In his epochal Rig-Vedic Repetitions Maurice Bloomfield reckons that about one-fifth of R̥gvedic verses can be considered to be repetitions (1916:4). The total of repeated whole verses rises to “not far from a third” when the variants of the other Vedas are taken into account (Bloomfield and Edgerton 1930:11).

Bloomfield of course sees that shorter phrases of noun and adjective, of verb and subject or object, and of local particle and noun are even more frequent than the repeated whole verses: “Set phrases, groups of two or three words—what Bergaigne used to call formulas3—are, as every Vedist knows, the commonplace of Vedic technique” (1916:xiv); “It will be seen that

1 This sketch arose from the research project “Familiengrammatik des R̥gveda” at the Indogermanisches Seminar of the University of Zürich, funded from May, 2006, to August, 2009, by Merbag AG, Zug, and by the Swiss National Science Foundation, Bern. I thank both of these far-sighted organizations for their support. The complete version, with full material, will appear as Chapter 6 of my R̥gvedic Family Grammar(forthcoming). R̥gvedic translations are those of Jamison and Brereton 2014.

2 Bloomfield similarly notes, “They read like two essays on the same theme, written by the same author, in two slightly differing moods” (1916:12).

3 Abel Bergaigne frequently did use the term “formula” in the general sense of “text-passage” or “statement,” but without regard to repeated word-groups (1878-83).


repetition of two or more consecutive words is an established feature of R̥g-Vedic composition” (4). He therefore distinguishes between “important, word-for-word repetition,” that is, of entire verses, and these “partial, less important repetitions” (4-5, 8-12). Of the latter he is remarkably scornful: “mere collocations of two or more consecutive4 words,” “merely consecutive words,” “mere groups of words or set phrases” (3-4); “unimportant, formulaic, and hap-hazard . . . expressions” (9); “conventional thought and mechanical utterance” (21); “A great many of the repeated passages consist of commonplaces, or are mere formulas” (22). He sees the inflection of a formula as an “unimportant stylistic or metrical accident” (9).

Bloomfield’s disdain for the “partial repetitions” follows naturally from his focus on repeated whole verses. This disinterest keeps him from rigorously analyzing these shorter repetitions, so that he has no way to decide whether the ten variants of 8.56.5c, agníḥ śukréṇa śocíṣā (“Agni with (his) blazing flame”), are modifications or different formulas (1916:9).

Of a R̥gvedic reverse concordance that Bloomfield created using the original cut-and-paste technique (1916:xvii, 2-3, 11), only the collection of 1,675 repeated cadences ever saw print (1916:653-74). This did suffice to prove that repetitions are far more frequent at the ends of verses than at the beginnings (1916:11).5 Bloomfield saw that in order to study the “partial repetitions,” that is, the formulas, even ab initio and a tergo concordances together would not suffice; instead something far more laborious, a “word-for-word concordance,” would be necessary (1916:3-4, 13); with Lubotsky’s work (1997), this dream has now become a reality.

2. Formulas in Homer and the R̥gveda

Less often cited than Milman Parry’s classic definition of the formula, “a group of words which is regularly employed under the same metrical conditions to express a given essential idea” (see A. Parry 1971:xxxii), but perhaps a better summary of his viewpoint, is the following: “The diction which is needed for making verses orally . . . is made of a really vast number of word-groups each of which . . . expresses a given idea . . . and fills just the space in the verse which allows it to be joined” (M. Parry 1971d [1930]:270). Both versions hold up well for the repeated noun-epithet formulas on which they are based, but less so when it comes to other types of repeated word-groups. A description of Parry’s work on formulas and oral poetry must be omitted here, but two general characteristics which he ascribes to formula-systems do need to be addressed.

2.1 Formular Economy

The avoidance of metrical doublets, known as formular economy or thrift, comes about because alternatives are needless in oral poetry. Parry admits openly and often that formular economy entirely obviates meaning (as regards the epithets at least): “one expression is useful in

4 Since Bloomfield is so insistent on this point let it be said that R̥gvedic poetic formulas are often split and over a fifth is inherently discontinuous.

5 A condition which holds true for Homer as well.


composition; equivalent expressions add no further advantage” (1971b [1928]:175)—except, of course, the advantage of differences in meaning being possible.

However Parry does in fact find a great many metrically equivalent formulas (1971b [1928]:173-89). A few he is able to explain away by analogy with other formulas or by truncation.6 But outside of the name-epithet systems, the doublets or “breaches of economy” are even more frequent.7 Friedrich concludes that between a fifth and a third of formula-systems present such breaches, and sometimes more than half. Formular economy is thus reduced from a principle to a tendency (2007:65, 140).

For the R̥gveda the notion of formular economy is utterly otiose; in its simpler measures the choice of metrically equivalent epithets is considerable, as for the two main deities:

Indra: śatakratu- = śacīpati-, kratumant- = harivant-, gopati- = satpati-, pūrbhid- = vajriṇ-, śakra- = śūra- = ugra-.

Agni: viśvavedas- = jātavedas-, ūrjā́m páti- = vaiśvānara-, havyavah- = viśvavid-, subhaga- = atithi-.

2.2 Formular Extension or Density

Parry’s teacher Antoine Meillet taught that Homer was entirely formulaic (1923:61), and Parry implies this as well (1971b [1928]:80, cf. 8-9, 21). Based on an analysis of fifty verses, and having loosened his definition of the formula to include parallel phrase structures,8 he concludes that formulas occur “one at least to every verse or so” (1971d [1930]:312). Reducing the sample to fifteen lines and using the same liberalized definition of the formula, Albert Lord reckons with “well over 90 per cent” of that text being formulaic (1960:144). Exiguous as they are, these samples have given rise to a “dogma of the 100% formularity of Homer” (Finkelberg 2004:245, cf. 236).

It took decades for objections to be raised. Arie Hoekstra opines that “the supposition that Homeric poetry is wholly formulaic is at all events unprovable (if not entirely unsound)” (1965:16). Joseph Russo notes the “surprisingly limited scope of these analyses . . . on which Parryan orthodoxy of 80-90 per cent is based” (1976:40). Although Brian Hainsworth agrees that only one verse in ten may be totally free of formulas (1968:16-17), he still finds the frequency of non-formulaic material to be “disturbingly high” (1962:66) and that “a large part of [the Iliad’s] diction is not formular in the strict sense” (1993:4, 17), estimating total formulaicity to be no higher than “from one-third to one-half of the total” (1964:164 and 1968:16-17, 131;

6 “A formula contained in a more complex formula . . . , formulae preserved because of their presence in more complex formulae” (M. Parry 1971b [1928]:180-81).

7 See Hoekstra’s index under “equivalents” and “thrift” (1965:167, 171); surprisingly skeptical is Hainsworth 1968:7 and 1993:24-26.

8 Such as, in an extreme example, δῶκεν ἑταῖρῳ and τεῦχε κύνεσσιν. These have been called “sentence or phrase patterns” (Hainsworth 1968:16-7, 41-42, and 1993:9-10) and “structural formulas” (Russo 1966:217-40). This definitional shift has not escaped criticism; see Hoekstra 1965:11-2, 15-16, 24-25; Hainsworth 1968:16-17; Russo 1997:242-46, with n. 19.


followed by Finkelberg 2004:245). Naturally the formulaic density can vary from place to place: Hainsworth (1968:110-12) contrasts the higher formulaicity of a battle scene with the lower formulaicity of a lament.

For the R̥gveda the question of formular density has been asked only in terms of the repeated whole verses. Bloomfield found these to constitute a fifth or more of the text, whereas the versus iterati that have fascinated Homerists since Aristarchus of Samothrace make up a third.9 The higher frequency of repeated whole verses explains why the Homeric type-scenes tend to pale in a way the R̥gveda never does. Conversely, due to its shorter verses and more limited subject matter, repetition in the R̥gveda can at times approach a hypnotic incantation in a way that Homer never does.

Although Homer repeats more whole verses, the R̥gveda preponderates in a specific type to which Parry (1971e [1933]:376-90) drew particular attention due to its usefulness in oral-poetic composition: verses containing exactly one sentence, that is, whose metric and syntactic borders coincide. In the Iliad such coterminous verses are one in ten, in the R̥gveda one in four (Dunkel 1996:206). The lesser use of one-verse sentences makes Homeric poetry flow more continuously.

The R̥gveda and the Homeric epics, both at least half formulaic, are quite comparable in bulk as measured by lexemes and syllables:

R̥gveda:39,67610 (8-12 syllables) 164,76611 9,89112   395,91513
Iliad + Odyssey:27,850 (12-18 syllables)198,837149,893403,00015 (?)

Of course, the non-formulaic (Parry’s “untraditional” and “unschematized”) language is no less important than the formulas, as its underived and unique expressions are crucial for the investigation of poetic originality.

9 Schmidt (1885:viii) counts 9,253 versus iterati, including minor variants, out of the total of 27,850 verses; see also M. Parry 1971b [1928]:8, n. 2.

10 This is based on the text of Van Nooten and Holland 1994.

11 As counted in the “Familiengrammatik des R̥gveda” project (see above, footnote 1).

12 This is the number of entries in Grassmann’s Verzeichnis der Wörter (1873:1690-1739), including the particles (358 of the total).

13 In the metrically restored text of Van Nooten and Holland 1994, as counted in the “Familiengrammatik des R̥gveda” project (see above, footnote 1).

14 This and the number of lexemes are according to the Thesaurus linguae graecae, available at

15 Estimated by reckoning 27,800 x 14.5 (the average between 12 and 17).


3. Formular Flexibility in Homer

Parry’s definition of the formula does not expressly exclude nominal and verbal inflection, but these concern him only insofar as they lead to hiatus or brevis in longo (1971b [1928]:68-74 and 1971c [1928]:197-201). The idea that Homer’s poetic formulas are basically fixed in form was rejected by Bryan Hainsworth (1962, 1968, and 1993) and Arie Hoekstra (1965) in favor of the view that the Homeric formula was flexible in various ways. Hainsworth’s final panoply of modifications includes change of word order, movement, inflection and suffixal variation, expansion, and separation (including enjambment); these can apply concurrently. He estimates between a third and a half of formulas to be flexible (1968:118-19, 122). In spite of all these types of modification, “the word-group persists” (Hainsworth 1993:26).

The effect is to break Parry’s intimate link between form and meter. Far from being the ultimate explanation for all formulaic usage, the meter is now just a framework over which the supple formulas disport themselves.

The reaction to this development has varied from acceptance, active or tacit,16 to “a confused state” (Russo 1997:250, cf. 242, 252), “general bewilderment,” and even to “a major crisis . . . and a defensive, if not apologetic, attitude” so extreme that publication in this field has “sharply decreased” (Finkelberg 2004:244-46).

As regards the R̥gveda there is no such controversy, since no overly stiff definition of the formula—or any definition at all—exists to react against. When one is put in practice, the formulas turn out to be even more mobile and flexible than Homer’s.

4. The Advent of Writing 

In India writing remained unknown until long after the completion of the authoritative saṃhitā-text (perhaps around 600 BCE). Its first appearance there in any form was the Aramaic script, brought by the Persian Achaemenids after 500 BCE. Over the centuries this served as the basis for the Brāhmī and Kharoṣṭhī scripts, first attested in King Aśoka’s Prākrit rock inscriptions of c. 250 BCE (see Salomon 2003:87-89, 92-93). Given this chronology, the R̥gveda cannot possibly have been composed, collected, and edited in any manner other than orally. The earliest preserved manuscripts of the R̥gvedadate from about 1350-1450 CE; they are practically irrelevant as regards its transmission.

While Parry’s guslari mostly still “re-created” orally, as they were illiterate (Lord 1960:20; Kirk 1962:84),17 in Greece the earliest rock graffiti and vase inscriptions are practically contemporaneous with the time assumed for “Homer,” about 750-700 BCE. Homer’s ignorance of writing has been the communis opinio since Friedrich Wolf’s 1795 Prolegomena ad Homerum, but since the 1950s the possibility has repeatedly been suggested that the proto-Iliad might have been written on skin or papyrus, either by a scribe (“oral dictation”) or by the singer

16 See the surveys by Windelberg and Miller 1980:29-50; Russo 1997:238-60 and 2011:296-98.

17 In fact becoming literate notably worsened their style (Friedrich 2007:138 n. 223).


himself (“oral autography”; Lord 1960:124, 129).18 This does not change the fact that Greek epic reached the level which made Homer possible over many generations without the use of writing.

The delay between the end of composition and the advent of writing was centuries long in India, but practically nonexistent in Greece. The period from the end of composition until the use of writing to record our texts differs even more between the two societies. Despite this, the oral styles of both the R̥gveda and Homer are still recognizable as such.

5. The Process of Canonization

Whatever “Homer himself” may have done, in Greece writing was early on felt to be necessary, whether due to continued poetic creativity or to less-than-perfect memory; acceptance was quick. Within two centuries Peisistratos not only needed, but was able to collect numerous official or approved texts from other municipalities. But after Śākalya’s saṃhitā the R̥gveda was not transcribed in writing for over a millennium.

The reason for the indifference to writing in India is the sheer quality of the brahmanic oral transmission, which prevented any variation. To this day the Vedic-Hindu tradition rejects any dependence on writing, just as did the Roman pontificesand the Gaulish Druids (Watkins 1976:107-08).19 Yet in contrast to the almost total loss of the latter’s hymnals, brahmanic misography has not affected the text of the R̥gveda in the slightest; as the most important possession of the priestly caste it has been transmitted with a rare exactitude, providing what has been called “a tape-recording of what was first composed and recited some 3,000 years ago,” a “snapshot of the political and cultural situation” which is “faithfully preserved, equivalent to inscriptions” (Witzel 1995a:91; see Bronkhorst 2002:797-99 and 2016:163-67). Due to this flawless mnemonic transmission the first written text, whenever and wherever it was made, was practically an irrelevance.

6. The Genesis of the Texts’ Present Form

The present forms of the texts were affected by both political and philologic factors in both societies. The earliest pre-R̥gveda, consisting of the kernels of the family books (2-7) and the Soma book (9), was created at the time when the latest R̥gvedic hymns were being produced, during the linguistic period of the Atharvaveda and the non-R̥gvedic mantras, perhaps around 1000 BCE. Witzel ascribes this to the mythologized King Sudās (or his successors), standardizing the text in order to consolidate the Pūru and Bharata peoples after his victory in the

18 The pro-writing arguments of Lesky and Erbse are summarized by Heubeck: “The Iliad and Odyssey . . . could not have been created at all without the aid of writing” (1988:12). See further Hainsworth 1968:2 n. 2; Burkert 1995:147-48; West 2011:9-11.

19 Farmer et al. (2004:44, 48) suggest that the Indus Valley culture deliberately embargoed imports bearing the marks of this pernicious practice.


Ten Kings’ Battle.20 The kernels of the composite books (1, 8, and 10) were added during the period of Yajurvedic prose; Witzel ascribes a pre-R̥gveda with ten books to the mythologized King Parikṣit, wanting to unite the “first Indian state.”21 All the books received further additions during the period of Brāhmaṇa prose.22 After a spell of anonymous editorial activity23 at the start of the Sūtra period (perhaps around 600 BCE) emerged the grammarian Śākalya’s saṃhitāpātha or “connected text” of 1,028 hymns and almost 40,000 verses, unchanged by a syllable since.

In Greece the creation of an unprecedentedly long and excellent proto-Iliad, perhaps about half of its present length, is ascribed to an Ionian Homer24 of around 750-700 BCE. This beloved text was subsequently expanded in various ways, leading to controversy at the competitive recitations of Homeric poetry at the Panathenaic festival. As a result, the Athenian tyrant Peisistratus of the sixth century BCE is said to have made a first standardization, collating the various texts κατὰ πόλεις, and to have produced an authoritative, translocal edition written in the Old Attic alphabet. In the following centuries new additions continued to be made (Atticisms, wrong word-divisions, and variants favoring particular groups), and the transliteration into the Ionic alphabet introduced metrical irregularities. As a result, textual criticism was found to be more necessary than ever, and the work of generations of grammarians of the third and second centuries BCE culminated in the Alexandrian edition of Aristarchus of Samothrace with its 28,000 verses. This text underwent considerable distortions in late antique and medieval times.

In sum, the canonizations of the Iliad and the R̥gveda involved surprisingly parallel processes:

  • An unknown agent collected the favorite oral compositions of a long poetic age into an unprecedentedly massive text (the kernels of books 2-7 and 9; the proto-Iliad).
  • The largely anonymous compositions were ascribed to specific males, partly invented (the traditional r̥ṣis of the Anukramaṇī; “Homer”).
  • The beloved text was expanded in various ways, leading to local differences (composite books; expansions of the epics).
  • An ambitious leader codified the collection so as to reduce controversies (the early ten-book R̥gveda; the Peisistratean recension).
  • Additions continued to be made as orality began to give way to simple reproduction.
  • The continued variation and increasing difficulties of comprehension called into being dedicated 

20 Witzel calls this the “Bharata collection” (1995b:337-38 and 1997a:290).

21 Witzel calls this the “Kūru collection” (1992:616, 1997a:261, 264-65, 286, and 1997b:35-36, 38, 40.

22 Namely 10.85-191, the Vālakhilyam, the maṇḍalar intrusions of books 3-5, and Oldenberg’s Anhangslieder.

23 Oldenberg’s “orthoepic diaskeuasis” was a misometric modernization, which transformed verse into prose by an inconsistent application of the much later sandhi rules of classical Sanskrit (whence saṃhitā), as if the goal were a “Zusammenpressen des vedischen Textes auf die möglichst geringe Silbenzahl” (Oldenberg 1888:461).

24 This has been suspected of being a professional or stage name, as seems to be the the case with Hesiodos, Stesichoros, as well as other Greek poets, and probably with R̥gvedic r̥ṣis such as Br̥haduktha (“Having high songs”) and Śrutavid (“Tradition-knower”).


philologists (the orthoepic diaskeuasts; Alexandrian grammarians), who eventually produced a definitive edition (Śākalya’s saṃhitāpatha; Aristarchus’ text).

7. R̥gvedic Repetitions: Non-Oral Approaches

Theodor Aufrecht’s explanation for the innumerable repetitions is epigonality: that they are mere remnants of a vanished poetry, “mere relics of an older or original Veda (jyeṣṭhám bráhma), . . . attributable less to direct imitation or unconscious reminiscence of the actual thing, than of what used to be” (1877:II, xii).

Maurice Bloomfield champions the epigonal point of view even more strongly. The repetitions he judges by modern literary standards, and his judgment is far from approving: “Vedic literary production is often in a high degree imitative and mechanical. The poets or priests, more or less consciously, fell into habits of expression such that entire lines . . . and considerable sequences of words . . . show much similarity” (1916:vii). Most of the repetitions are “literary or historical in nature” and indicate an “imperfect sense of literary proprietorship” or “plagiarism” among the r̥ṣis (19). The high degree of repetition is the result of “reciprocal assimilation” (20): “R̥gvedic repetitions are often due to more or less conscious imitation” (634). He does not speak of r̥ṣis imitating r̥ṣis, but of hymns and stanzas imitating other hymns and stanzas: “A pāda, stanza, or strophe . . . may imitate another without directly repeating its words, but in the manner of a paraphrase” (12). The Vālakhilyas are “entire hymns that are consciously imitative” (13). Correspondingly Bloomfield faults the Anukramaṇī for “find[ing] it in its heart to assign, with unruffled insouciance, one and the same verse to two or more authors, or to ascribe it to two or more divinities” (634). Of course, nothing is more fundamental to oral poetry than a common stock of formulas.

The term “orality” he uses only in reference to transmission, not composition, having “little doubt that this oral tradition [of transmission] was supported at a comparatively early time by written tradition (see AV 19.72)” (1916:vii).

Pavel Poucha puts a positive spin on the repetitions: “The old poets considered quoting from others to be honorable rather than a lack of originality” (1942:250). He thinks that the repeated verses’ assonant figures of style make them easier to learn and thus more frequent (257-69). At the same time, the fact that only 11.2% of hymns are free of repeated whole verses shows the “lack of proper literary training of the composers” (250).

In a surprisingly influential footnote, Albert Lord briefly dismisses any relevance of “sacred texts which must be preserved word for word, if there be such” for the study of oral poetry (1960:280 n. 9), on the ground that the Vedic hymns have long been fixed and not “re-created” in performance. Lord is using the R̥gveda’s mode of transmission as a straw man to avoid the question of its method of composition. He does not deign to mention that text’s extremely high degree of repetition as established over 40 years earlier by Bloomfield. But this sentiment was to keep research on R̥gvedic orality in the closet for a quarter century.

Jan Gonda (1975:193-97, 221-30) discusses formulas, repetition, refrains, similarities, parallelisms, and variation in the R̥gveda at considerable length without suggesting anything new. He accepts everything anybody has previously said except for those denying the presence


of alliteration (224). However his brief mentions of oral poetry and formulas show no understanding of its improvisational nature (28 n. 26, 74-75, 221), as in his reference to “the works of predecessors which they had memorized” (193). Like an oral poet himself, Gonda repeats his predecessors’ literacist formulations, speaking of “an imperfect sense of literary proprietorship” (193) and of “the stereotyped literary form of the R̥gveda and the problem of recasts and borrowings” (28), and averring that “the earlier poets had exploited these themes so thoroughly that nothing was left for their successors but to follow in their habits” (194, approximating Parry’s view on originality). Despite his oft demonstrated interest in linguistic repetition, he does not mention its connection with performance in public. This is a distinct step backward from his earlier position (Gonda 1959a), perhaps due to Lord’s portentous footnote.

Jack Goody (1985:7-17 and 1987:110-21) thinks that the Vedas are too vast and too consistent to have been composed and transmitted orally, since oral poetry from all over the world is characterized by widespread textual inconsistencies. Once again: the perfect transmission has no implication for the method of composition.

Michael Witzel (1997a:258-59) uses the term “oral” only in the sense of non-written, not in that of formulaic and improvisational composition. Elsewhere he states that the R̥gveda was “composed in a traditional and complicated poetic language like the Iliad” (Witzel et al. 2007:477, cf. 448, 475), but still speaks of the Vedic r̥ṣis’ “shamelessly copying” each other (448) and characterizes many Vedic hymns as “stereotyped” (451). He uses the term “formula” only in an untechnical, pre-Parry sense: “traditional formulas, figures of speech, epithets”; “pre-existent formulas, mobile components, epithets and kennings”; about the repetitions he says, “the poets often borrow even from their predecessors” (446-48). For Witzel, as for Aufrecht, the R̥gveda is only an “afterglow” of Proto-Aryan and Proto-Indo-European poetry (449).

 Jared Klein has devoted over twenty-two articles (listed in Klein 2012:191-201) to stylistic repetition in the R̥gvedawithout ever mentioning oral-poetic formulas.

The striking paucity of work on R̥gvedic orality is shown by the lack of a single paper on this topic in the thirty-four previous volumes of Oral Tradition.

8. Vedic Orality: Scholarly Acceptance 

Parry’s ideas were accepted by Jan Gonda. He states that both Homer and the R̥gveda are “traditional” in nature, and “improvised” by “oral poets” who were “neither free in their choice of words nor original in their invention: these very formulas and fixed expressions set them bounds and forbade them the search for an individual style” (1959a:254); the traditional oral-poetic formulas exist “to make it easier for the poet to compose as well as for the audience to listen” (1959a:29, 31, 254). But for both texts he rejects Parry’s idea that the epithets serve only metric purposes (see below, section 13.4).

But Lord’s dictum against R̥gvedic orality the following year caused Gonda to abandon his acceptance in 1975, and in fact stifled any discussion of this topic until 1976, when Paul Kiparsky finally dared to contradict him: “Lord excluded the Vedic literature from oral poetry by fiat, in reserving the term ‘oral poetry’ for poetry composed during performance. This would make the most important thesis of Lord’s book true by definition” (Kiparsky 1976:101).


Kiparsky sees the genesis of both the R̥gveda and Homer as “the collective elaboration of a fixed text out of a tradition of oral poetry . . . [by] a bardic guild” by means of “a gradual fixation of the text over several generations of continuous recitation by a family or guild of singers,” “a fluid oral tradition ‘freezing’ into an absolutely rigid shape,” “a gradual jelling of an initially loosely connected body of poetry which was gradually added to and reorganized” (102-04). In charmingly idealistic contrast to the usual view of Vedic society as riven by tribal jealousy and feuds, Kiparsky suggests that “what the singers probably did was to sit together and perform things for each other . . . and gradually a stable version was worked out” (in Stolz and Shannon 1976:116), that is, the collection arose as the result of Vedic Woodstocks. These would have fostered mutual borrowing and thus contributed to the homogenization of the R̥gvedic poetic language.

In his response Calvert Watkins felt free at last to admit that “the formulaic character of the composition of the Vedic hymns is apparent in virtually every mantra” and, one imagines with a sigh of relief, to “welcome Kiparsky’s principled inclusion of Vedic poetry within the universal discourse of this conference [on oral poetry]” (Watkins 1976:107-08). He went on to reject Parry’s phrase, “regularly employed under the same metrical conditions” (109). In 1995 Watkins sees the formula as “a verbal and grammatical device for encoding and transmitting a given theme . . . . Theme is the deep structure of formula” (1995:17). He repudiates Lord’s dictum again (18), and also the phrase “group of words” in Parry’s definition of the formula by accepting single words as formulas (17). He operates as a matter of course with formulaic modification and lexical renewal (10, 15).

Applying Parry’s statistical measures of relative orality—frequency of enjambment as a whole, frequency of coterminous verses, and frequency of necessary and violent enjambments—to the R̥gveda, George Dunkel finds its style to be distinctly more oral than that of the Iliad (1996:204-06). Elsewhere he uses formulaic theory to resolve some longstanding syntactic controversies. The alleged deletion or “gapping” of repeated preverbs and verbs in Vedic and Homer is often due to the reuse of formulas outside their original environments (1978:14-26). Formular truncation has led to oddities such as the seemingly conjunctive use of emphatic and local ā́ and missing endings as in návyasā vácas (1982a:89-102) and to the so-called inverse ca (129-43).

Stephanie Jamison allows that the R̥gveda was “composed entirely orally and transmitted entirely orally” but still follows Lord in taking it as “a type of oral composition very different from” Homer’s because “it was not an anonymous body of infinitely variable verbal material (re-)composed anew at every performance” (Jamison and Brereton 2014:I, 14). But neither is our Iliad! And hers is a perfect description of the long period of free oral composition which preceded the fixation of the saṃhitā.

Outside of the Veda, Indology has long since recognized elements of oral composition in the classical epics,25 and its traces have also been found in the R̥gveda’s closest linguistic and cultural relative, the Gāthās of Zarathustra (see Skjærvø 2012).

25 See the Purāṇic and Epic Bibliography at


9. The R̥gvedic Poetic Formula

Since Bloomfield’s 1916 work much has been accomplished in R̥gvedic linguistics, mythology, and society, but no more large-scale research on its poetic formulas has been undertaken. Here we shall apply the methods of formulaic analysis developed in Homeric studies to the notoriously repetitive R̥gveda. The need for this became clear during an investigation of linguistic differences between the six great book families. We defined a familectally distinctive feature as one which recurs at least thrice in one family book and nowhere else. The “Familiengrammatik des R̥gveda” project (see above, footnote 1) collected over a thousand recurring pairs of designators of possible interest, which I then filtered so as to arrive at the 177 candidate formulas for familectal distinctiveness that are analyzed below. However the following discussion is not limited to these, since during analysis countless related formulas were also examined in the same way.

9.1 Definition 

A R̥gvedic poetic formula is a repeated, semantically unified word-group. The words’ position, form, function, and syntactic relation are irrelevant.

By “repeated” is meant occurring in the R̥gveda thrice or more. Repetition is the first criterion of formularity. But although necessary, recurrence is not a sufficient condition; this is shown by the recurring chance collocations, that is, word-groups that are repeated without being formulas.

By “word” or “element” is meant “designator” (noun, adjective, or verb): as is traditional, we ignore the formators (particles, primary adverbs, pronouns, and the like).

“Group” reflects the fact that cooccurrence is the second criterion for formularity. Single words cannot be considered to be formulas.26 Although the project originally searched only for recurring designator-pairs, analysis showed that many of these belonged to longer formulas, so that our candidate formulas can be six words or even an entire stanza in length;only 44% of the candidate formulas are limited to two words.

By “semantically unified” is meant that despite all formal modifications the elements continue to “express a given essential idea” (so Parry; cf. Aufrecht’s “express the same ideas in a somewhat different style”; 1877:II, xi)However defined, this unity or identity is the third and final criterion of formularity. Hainsworth speaks of a high “degree of mutual expectancy” between the elements:

  • “The use of one word created a strong presumption that the other would follow” (1968:35-36) in a “certain formulaic word association” (61).
  • “Formulas are simply groups of two or more words that are associated with each other” (1993:18).
  • “The word-group persists in spite of declension or conjugation, changed localization, expansion,

26 On the stylistic repetition of individual words see Gonda 1959b and the twenty-two articles mentioned by Klein (2012:191-201), which have since doubtless been joined by others. Neither author mentions poetic formulas.


or shortening. . . . The formular link may even survive enjambement . . . ” (1993:26-27).

Of course, “mutual expectancy admits of infinite gradations” (1968:41).

A more formal way of saying that the meaning remains unchanged is distributional: since the modifications preserve the cooccurrence restrictions (or privileges of occurrence) of the formula’s elements, they can be seen as transformations of the formulas (Harris 1957), as paraphrases of their reports (Harris 1970:612-92).

Since a formula’s identity is not syntactically defined, its variants need not always be construed the same way, but they will talk about or mention the same thing. If the formula’s elements should happen to cooccur without expressing the same essential idea, this is considered to be not a repetition, but rather a chance collocation or a different formula.

9.2 Formula and Meter in the R̥gveda

A formula can fill a whole verse, be shorter, or be longer. A formula’s boundaries practically never differ from the metric ones, both between the verses and within them.27 While synchronically the formulas seem tailored to fit the meter, historically they may have played a role in creating it.28 The transfer of formulas between meters often induces reduction, extension, enjambment, new boundaries, and so on.

Metrical pressure is weaker in the R̥gveda than in Homer, since only the number of syllables is crucial; except for the cadence, their quantity is less important. This is illustrated by the relative rarity of completely artificial formations when compared with Homer, who has numerous forms which are found in no real dialect. Little dialect mixture can be registered and very few hyperforms.29

It often happens that the opening and the cadence of a trimeter verse are filled by four- or five-syllable formulas. The intervening break can then be filled either by expanding one of the formulas or by inserting a link-word (by definition not a formula).

8.100.4aayám asmijaritaḥ# páśya mehá
6.9.4aayáṃ hótāprathamáḥ# páśyatemám
6.5.6dtáj juṣasvajaritúr# ghóṣi manma
2.11.1aśrudhi havamind(a)ra# mā́ riṣanyaḥ

27 That is, between the opening, the break if there is one, and the cadence. In Homer, on the other hand, “the relation between colon and formula is so obscure that as an element of composition it may well be irrelevant” (Hainsworth 1968:20).

28 According to Gregory Nagy “predictable patterns of rhythm emerge from favorite traditional phrases” and “formula generates meter” (1976:251-52; see 1990:18-35).

29 The only hyperforms due to metrical pressure are unjustified disyllabic ā, anti-Sievers forms like áśua– for áśva-, and the first singular active subjunctive ending –āni.


10. Formulaic Flexibility in the R̥gveda

Although Bloomfield considered the inflection of a formula to be an “unimportant stylistic or metrical accident” (1916:9), only one-sixth of our formulas are completely fixed; the rest are flexible to some degree, as illustrated by the formula, “to smash the demons”:

6.16.29c jahí rákṣāṃsi sukrato
9.17.3c vighnán rákṣāṃsi devayúḥ
9.49.5b rákṣāṃsi apajáṅghanat
9.63.29a apaghnán soma rakṣáso

or by the formula, “Soma lengthens (our) lifetime”:

8.48.4d prá ṇa ā́yur jīváse soma tārīḥ
8.48.7c sóma rājan prá ṇa ā́yūṃṣi tārīr
8.48.10cd ayáṃ yáḥ sómo ní ádhāyi asmé / tásmā índram pratíram emi ā́yuḥ
8.48.11cd ā́ sómo asmā́ aruhad víhāyā / áganma yátra pratiránta ā́yuḥ
9.80.2cd maghónām ā́yuḥ pratirán máhi śráva / índrāya soma pavase vŕ̥ṣā mádaḥ
10.107.2d vāsodā́ḥ soma prá tiranta ā́yuḥ.

Bloomfield classified the types of variation among repeated verses under two headings: “Metrical variations as results of addition or subtraction or verbal change in repeated pādas” (1916:523), involving changes in meter, and “Verbal variations of repeated pādas: lexical and grammatical” (548), involving inflection and lexical substitution. This is a useful first step, but when we change the focus from repeated verses to formulas, it proves inadequate. Expanding Hainsworth’s system we have arrived at the following nine types of modification which have proven to be both necessary and sufficient to account for the flexibility of all formulas we have seen.

Flexibility is of two basic types: formulaic modification and lexical substitution. “Modification” encompasses any change in a formula’s shape or structure, but not its word inventory. Any number of modifications can apply concurrently.

The Types of Formulaic Modification in the R̥gveda

Change of formChange of meaning
Change of positionLengtheningShortening
1. Inflection
2. Syntactic transformation
3. Movement6. Extension: at the margin8a. Truncation
8b. Reduction
9. Metanalysis: change in structure or meaning, not form


4. Inversion7. Split, Insertion
(internal lengthening)
8c. Ellipsis                                   
5. Enjambment

10.1 Inflection

Within a paradigm the number of syllables may remain unchanged:

ahan—hanti—hanas—jahi—hatam             jaritā́—jaritar—jaritúṣ;

it may be increased:

thematic nom. pl. –āsas beside –ās              thematic instr. sg. –ena beside –ā;

or it may be allowed to vary:

śrudhi, śr̥ṇudhi               yaja(sva)            piba(ta) r̥túna                 paramé vyòman(i).

Derivation within an inflectional category, that is, the production of new stems, belongs here as well. It usually involves suffixes:

  • Suffixal variation:

                   Dadhikrā́(van)-, Médh(y)ātithi- (personal names)
                   r̥jipyá-, r̥jīpín- (epithet of an eagle)

  • Conversion to paradigmatic participles and verbal adjectives, as when the thrice-attested formula úd eti sū́ryas (“the sun rises”) is transformed into udyántaṃ tvā . . . sūrya (thrice) or the locative absolutes sū́ra údite (9x) and sū́rya udyatí (once).
  • Change to verbal secondary stems, as in the formula, “to strike the demons”:

                        9.63.29a apaghnán soma rakṣáso     beside the causative
                        9.49.5b rákṣāṃsi apajáṅghanat

Inflection and derivation occur together in “to know the ascent of heaven”:

            4.8.4c vidvā́m̐ āródhanaṃ diváḥ      beside 
            4.7.8d vidúṣṭaro divá āródhanāni

10.2 Syntactic Transformation

This category encompasses diverse types of modification.


10.2a Subordination

This occurs in the formula, “the two go to the clan”: 

7.73.4a úpa tyā́ váhnī gamato víśaṃ no      but relativized in
7.69.2c víśo yéna gáchatho devayántīḥ      and causal in
7.74.1cd víśaṃ-viśaṃ hí gáchathaḥ

10.2b Passivization

Passivization of active verbs often involves the verbal adjective, as in the formula, “to prop apart heaven and earth”: 

6.44.24a ayáṃ dyā́vāpr̥thivī́ ví ṣkabhāyad     beside
6.70.1cd dyā́vāpr̥thivī́ váruṇasya dhármaṇā / víṣkabhite ajáre bhū́riretasā

It may also involve the gerundive, as in the formula, “to choose Agni as messenger”:

1.12.1ab agníṃ dūtáṃ vr̥ṇīmahe      beside
8.102.18bc agne dūtáṃ váreṇyam / havyavā́haṃ ní ṣedire 

10.2c Nominalization 

Verbs can be transformed into abstract nouns and infinitives, as when pan-R̥gvedic sutám piba / piba sutám (“drink the pressings!”) (7x, 4x) appears as sutásya pītáye (8x) and sutásya pītím / -íṣ (“a drink of the pressings”) (2.11.17d, 4.35.2b).

Verbs can also be transformed into agent nouns by suffixation or composition (with the above cf. somapā́- (12x)). A change of mood can cause another verb to be inserted:

8.84.3a nr̥̄́m̐ḥ pāhi # śr̥ṇudhī́ gíraḥ 
             “Protect the men! Hear the songs!”      but
2.20.3b sákhā śivó narā́m astu pātā́
             “Let him be a benevolent companion and protector of the men”

10.2d   Stem Composition

This is another type of nominalization, which obscures the first element’s syntactic relation to the second. The verb of the formula, “the stronghold-splitter . . . to make,” is adverbialized and nominalized, respectively, between:

8.61.8c ā́ puraṃdaráṃ cakr̥ma vípravacasa      and
8.61.10a ugrábāhur mrakṣakŕ̥tvā puraṃdaró      and
8.1.7c álarṣi yudhma khajakr̥t puraṃdara.


10.2e Verbalization 

When a root-compound occurs in variation with a finite verb, the nominal form need not necessarily be secondary. The phrase āródhana- divás (“the ascent of heaven”) occurs five times, but it is verbalized only once (āroháyanti diví); the compound śucipā́- occurs five times, but is verbalized only once (piba śúcim, “drink (it) pure”). The pan-R̥gvedic formula ádribhiḥ sutá- sóma- (“Soma pressed with stones”) occurs fourteen times, but it is verbalized only thrice:

4.45.5d sómaṃ suṣā́va mádhumantam ádribhiḥ
9.34.3b sunvánti sómam ádribhiḥ
9.107.1d suṣā́va sómam ádribhiḥ

10.2f Simile

A formulaic element may be transformed into a simile by the addition of a particle  meaning “like” (, iva, yáthā) without affecting the formula’s unity, as in the formula, “to cross hates (and) straits”:

6.2.1 = 6.14.6d dviṣó áṃhāṃsi duritā́ tarema
6.2.4d dviṣó áṃho ná tarati

An element is shifted out of a simile in the formula, “to be swollen like ghee”:

8.7.19b ghr̥táṃ ná pipyúṣīr, 8.12.13c ghr̥táṃ ná pipya, but
8.6.43b mádhor ghr̥tásya pipyúṣīm (“swollen full of honey and ghee”)

The next three modifications involve change in the elements’ position.

10.3 Movement 

A formula can move within a verse, as with “enjoy that!” and “lofty light”:

4.2.20b avocāma kaváye # tā́ juṣasva
6.5.6d táj juṣasva jaritúr # ghóṣi mánma
6.47.10d táj juṣasva # kr̥dhí mā devávantam

1.45.8c br̥hád bhā́ḥ bíbhrato havír
4.5.1b kathā́ daśemāgnáye # br̥hád bhā́ḥ


10.4 Inversion

Inversion of words has been recognized as an element of high style since the Greek Sophists; for the R̥gveda see Bloomfield 1916:7, 552-53. Limiting ourselves to contiguous words, we may cite as examples the formulas, “I invoke Agni” and “Drink of this!”:

1.1.1a agním īḷe puróhitaṃ
3.27.2a ī́ḷe agníṃ vipaścítaṃ

3.35.6 śasvattamáṃ sumánā asyá pāhi
5.43.3c hóteva naḥ prathamáḥ pāhy asyá

The formula, “go home!,” exhibits inversion combined with movement:

10.95.2c púrūravaḥ púnar ástaṃ párehi
10.95.13d párehy ástaṃ # nahí mūra mā́paḥ

The formula may contain more than two words, as in, “Deliver the singer from narrow straits”:

1.58.8c ágne gr̥ṇántam áṃhasa uruṣya
1.58.9c uruṣyá agne áṃhaso gr̥ṇántam

10.5 Enjambment 

The running on of a sentence into the next verse, that is, its continuation over a verse boundary, is one of the two fundamental deviations from coterminosity (the other being verse-internal placement of a sentence boundary). 16% of the candidate formulas are inherently enjambed, their elements never cooccurring within a single verse. The formula might be said to contain a verse boundary—which, like any other formulaic element, can be mobile. About the same proportion of our formulas are enjambed in more than one way, as in, “to convey the gods who wake at dawn toward”:

1.44.1cd ā́ dāśúṣe jātavedo vahā tvám / adyā́ devā́m̐ uṣarbúdhaḥ
1.44.9cd uṣarbúdha ā́ vaha sómapītaye / devā́m̐ adyá svardŕ̥śaḥ
1.14.9 ā́kīṃ sū́ryasya rocanā́d / víśvān devā́m̐ uṣarbúdhaḥ / vípro hótehá vakṣati

Maximal enjambment means continuing over an entire stanza, as in, “to call (on) Indra of a hundred resolves with praises”:

8.52.6cd vasūyávo vásupatiṃ śatákratuṃ / stómair índraṃ havāmahe
8.52.4 yásya tvám indra stómeṣu cākáno / vā́je vājiñ chatakrato
            táṃ tvā vayáṃ sudúghām iva godúho / juhūmási śravasyávaḥ


The next two modifications involve changes at the formula’s margins.

10.6 Extension

The extension of a formula is its lengthening by inflection or by adding elements.30

10.6a Juxtaposition

The simplest type of extension is juxtaposition or concatenation, that is, adding a word or a phrase at a margin, as in, “your most delightful favor”:

7.70.2a síṣakti sā́ vāṃ sumatíś cániṣṭhā      which is extended into 
7.57.4d asmé vo astu sumatíś cániṣṭhā      and then reused elsewhere:
7.70.5d asmé vām astu sumatíś cániṣṭhā.

10.6b Overlapping

Overlapping is the combination of formulas sharing an element (“word association”; Hainsworth 1962:65), as when the formulaic variants br̥hád arca (“to chant aloft”) and br̥haté arca (“to chant to the lofty one”) are combined:

1.9.10bc br̥hád br̥hatá éd aríḥ / índrāya śūṣám arcati

A new overlap may itself become formulaic, as when the following verses:

5.41.16a kathā́ dāśema námasā sudā́nūn (“how should we do pious service?”) and
7.14.1d vayáṃ dāśema agnáye (“we would do pious service for Agni”)

are combined and the combination then reused:

1.77.1a kathā́ dāśemā agnáye # kā́smai
4.5.1b kathā́ dāśemā agnáye # br̥hád bhā́ḥ.

Three formulas overlap when dūtá- páti- agne (“the messenger, the lord, O Agni”) (thrice), víśā́m páti- (“lord of settlements”) (10x), and viśā́m asi (“you are of the settlements”) (thrice) are combined in:

1.44.9ab pátir hí adhvarā́ṇām / ágne dūtó viśā́m asi

and again when the formulas han- rákṣāṃsi (“smite the demons”) (7x), sedha- rákṣāṃsi (“keep

30 This is unrelated to the Parryan sense of “extension” (section 2.2).


away the demons”) (thrice), and sédha- ámīvās (“keep away the diseases”) (twice) overlap and recur as a hymn-internal refrain:

8.35.16-8b hatáṃ rákṣāṃsi # sédhatam ámīvāḥ.

10.7 Insertion of a Split

A split into a discontinuous formula occurs when additional words are inserted. A short interruption can nonetheless be important, changing “I exist” to the copula “I am”:

8.100.4a ayám asmi jaritaḥ # páśya mehá
10.83.5 ayáṃ te asmy # úpa méhy arvā́ṅ

Longer ones can be banal, as in, “the cooked within the raw”:

2.40.2c ābhyā́m índraḥ pakvám āmā́sv antáḥ
1.62.9c āmā́su cid dadhiṣe pakvám antáḥ

10.7a Maximal Split 

Maximal split of a formula, that is, over an entire stanza, often coincides with maximal enjambment, as when the inherently enjambed bisentential formula, “Come to [place-name], drink Soma like a thirsty [animal name]”:

8.4.10ab ŕ̥śyo ná tŕ̥ṣyann avapā́nam ā́ gahi # / píbā sómaṃ váśām̐ ánu

is split further into

8.4.3 yáthā gauró apā́ kr̥táṃ / tŕ̥ṣyann éti ávériṇam
         āpitvé naḥ prapitvé tū́yam ā́ gahi # / káṇveṣu sú sácā píba.

10.7b Inherent Discontinuity

Over a fifth of the candidate formulas are inherently discontinuous;31 here any previous contiguous version has fallen out of use. Even fixed formulas can be inherently discontinuous, as in, “great in might”: 

8.6.1a mahā́m̐ índro yá ójasā,  8.6.26c mahā́m̐ apārá ójasā, 8.33.8d mahā́ṃś carasi ójasā, and 1.9.1c mahā́m̐ abhiṣṭír ójasā

31 Hainsworth calls these “discrete formulae” (1968:91, 104).


Of course, inherently discontinuous formulas can be enjambed as well, as in, “Indra along with the Maruts drinks the Soma”:

3.51.7a índra marutva ihá pāhi sómaṃ
3.47.1ab marútvām̐ indra vr̥ṣabhó ráṇāya / píbā sómam anuṣvadhám mádāya
3.50.1ab índraḥ svā́hā pibatu yásya sóma / āgátyā túmro vr̥ṣabhó marútvān
8.76.4 ayáṃ ha yéna vā́ idáṃ / svàr marútvatā jitám / índreṇa sómapītaye
8.76.6 índram pratnéna mánmanā / marútvantaṃ havāmahe / asyá sómasya pītáye

10.7c Inherent Contiguity

Formulas may also be inherently contiguous, that is, unsplittable, as in repeated whole verses. But inherently contiguous formulas can also be enjambed, even in various ways, as in, “Viṣṇu strode out three steps”:

8.52.3c yásmai víṣṇus trī́ṇi padā́ vicakramá
1.22.18ab trī́ṇi padā́ ví cakrame / víṣṇur gopā́ ádābhyaḥ
1.22.17ab idáṃ víṣṇur ví cakrame tredhā́ ní dadhe padám

and in, “Indra puts the pressed Soma into his belly”:

3.35.6cd asmín yajne barhíṣi ā́ niṣádya / dadhiṣvémám jaṭhára índum indra
3.22.1ab ayáṃ só agnír yásmin sómam índraḥ / sutáṃ dadhé jaṭháre vāvaśānáḥ
3.40.5 dadhiṣvā́ jaṭháre sutáṃ / sómam indra váreṇyam / táva dyukṣā́sa índavaḥ.

10.7d Sentential Split

The most extreme type of formular split is that into two sentences. The mechanism is the insertion of either an additional verb (26x), as in, “to drive toward the good praise”:

8.34.1ab éndra yāhi háribhir / úpa káṇvasya suṣṭutím
8.8.6cd ā́ yātam aśvinā+ # ā́ gatam / úpemā́ṃ suṣṭutím máma

or of a verse-internal sentence boundary (32x), as in, “to sacrifice to the gods with this offering”:

7.17.3a  ágne vīhí havíṣā # yákṣi devā́n      beside
3.17.2c evā́néna havíṣā yakṣi devā́n.

Another mechanism of sentential split is the syntactic reassignment of an element to a preceding verb in, “O Indra, drink this Soma!”:

10.24.1a índra sómam imám piba, 3.32.1a índra sómaṃ somapate píbemám but
8.17.1ab ā́ yāhi, suṣumā́ hí ta / índra sómam # píbā imám


None of these syntactic splits affects the formula’s unity; if it did, the repetition would be invalid as a dissolution.

10.7e Subordination

Formulas can be split into subordinate and main clauses. Only twice is this by means of a particle (concessive ); the others are by relativization, as in, “Soma lengthens (our) lifetime”:

8.48.4d prá ṇa ā́yur jīváse soma tārīḥ      and
8.48.7c sóma rājan prá ṇa ā́yūṃṣi tārīr      beside
8.48.10cd ayáṃ yáḥ sómo ní ádhāyi asmé / tásmā índram pratíram emi ā́yuḥ  and
8.48.11cd ā́ sómo asmā́ aruhad víhāyā / áganma yátra pratiránta ā́yuḥ

Even two-word formulas can be be split by relativization, as in, “pressed Soma”:

9.107.1ab párītó ṣiñcatā sutáṃ / sómo yá uttamáṃ havíḥ

The formula vásavo juṣanta (“the good ones enjoy”) occurs thrice contiguously and twice split into subordinate and main clauses: once by , once by both relativization and vocativization.

7.5.6ab tvé asuryàṃ vásavo ny r̥̀ṇvan / krátuṃ hí te mitramaho juṣánta
5.3.10ab bhū́ri nā́ma vándamāno dadhāti / pitā́ vaso yádi táj joṣáyāse

10.8  Truncation, Reduction, and Ellipsis 

When longer and shorter variants coexist, it is not always clear whether this is due to extension or to reduction, as with, “to praise and sing to Indra”:

2.20.4a tám u stuṣa índraṃ, táṃ gr̥ṇīṣe      beside
8.65.5a índra, gr̥ṇīṣé u stuṣé32

When a variant loses marginal elements, we call it truncation; when it uses shorter allomorphs, we call it reduction, as in, “Become for us a giver of cows”:

3.30.21d asmábhyaṃ sú maghavan bodhi godā́ḥ      beside
8.45.19c godā́ íd indra bodhi naḥ

When a variant lacks some of the formula’s elements, we call it ellipsis.

In order to keep the assumption of ellipsis within reasonable bounds, we insist that a variant retain at least two elements of the full formula in order to count as a valid repetition. A

32 The first passage contains far-deictic u, the second, conjunctive u (see Dunkel 2014:II, 822).


maximal reduction down to the minimal two words is not infrequent, as when the inherently enjambed, six-word formula práti váraṃ jaritré / duhīyád indra dákṣiṇā (“May the honorarium yield milk for the singer according to his wish, O Indra”) (attested seven times) is reduced to dákṣiṇā duhīta in 2.28.8b.

10.9 Metanalysis 

This category of modifications involves change in structure without change in form.

10.9a Morphologic

 Morphologic metanalysis involves ambiguous endings. The formula br̥hád arca- (“chant aloft”) shifts between the first singular subjunctive in:

5.85.1ab prá samrā́je br̥hád arcā gabhīrám / bráhma priyáṃ váruṇāya śrutā́ya

and the second singular imperative in:

5.25.7ab yád vā́hiṣṭhaṃ tád agnáye / br̥hád arca vibhāvaso.

The form kánīyasas (“younger”) shifts between the genitive singular in:

7.86.6c ásti jyā́yān kánīyasa upāré 
             “The elder exists within the misdeed of the younger,” 

and the accusative plural in:

7.32.24ab abhī́ ṣatás tád ā́ bhara+ / indra jyā́yaḥ kánīyasaḥ
                 “Bring this greater (good) to those who are lesser.”

10.9b Semantic 

This form of metanalysis involves homonymic words. For example, padā́, the nominative-accusative plural of padám (“step”) in the formula, “Viṣṇu strode out three steps” (7x), as in:

8.52.3cd yásmai víṣṇus trī́ṇi padā́ vicakramá

shifts to the instrumental singular of pád- (“foot”) in:

6.59.6d triṃśát padā́ ny àkramīt
             “he trampled thirty with his foot.”


10.9c   Metanalysis of Syntactic Boundaries 

Sentence boundaries are not marked in the saṃhitā. This is usually innocuous because they almost always occur at verse end; in only two percent of verses are sentence boundaries shown by an accented verse-medial verb to be internal. Passages which contain the same words with and without an internal sentence boundary, such as:

3.17.2c evā́néna havíṣā yakṣi devā́n
             “So sacrifice to the gods with this offering”
7.17.3a ágne vīhí havíṣā # yákṣi devā́n
             “Agni, pursue them with the offering, sacrifice to the gods,”

raise the question whether the sentence boundary has been inserted or lost—whether a formula has been split or two formulas have been merged.

It is also possible for a sentence boundary to change its position without being formally marked, this being a true metanalysis:

10.27.24a sā́ te jīvā́tur # utá tásya viddhi
                 “This is your means of life. And know this!”

7.72.2cd  yuvór hí naḥ sakhyā́ pítryāṇi / samānó bándhur utá # tásya vittam
                 “For in you two are our ancestral companionships and common kinship.
                 Be aware of this!”

10.9d Metanalysis of Phrase Structure 

A loss of congruence need not affect the unity of the formula. This may be brought about by:

  • Vocativization, as in udyántaṃ tvā . . . sūrya (10.37.7cd) beside sū́ra údite (9x) and sū́rya udyatí (8.27.19c), or in tvā́ṃ citraśravastama (“thee, O with brightest fame”) beside tvā́ṃ . . . citrám (“thee, the bright”).
  • Inflection of an element: etā́vat- (“so much”) is attributive to sumná- (“goodwill”) in:

            8.5.27 etā́vad vāṃ vr̥ṣaṇvasū / . . . / gr̥ṇántaḥ sumnám īmahe      and
            8.49.9 etā́vatas ta īmahe / índra sumnásya gómataḥ

but possessive (“the goodwill of such a one”) in:

            8.7.15ab etā́vataś cid eṣāṃ / sumnám bhikṣeta mártyaḥ.

  • Transfer of an element to a neighbor: the parallelism of “accompanied by horses, cow, heroes” with shared referent in:


            7.41.7ab áśvāvatīr gómatīr na uṣā́so / vīrávatīḥ sádam uchantu bhadrā́ḥ      and
            7.75.8ab            nū́ no gómad vīrávad dhehi rátnam / úṣo áśvāvad purubhójo asmé

is broken up in

            9.63.18 ā́ pavasva híraṇyavad / áśvāvat soma vīrávat / vā́jam gómantam ā́ bhara.

11. Lexical Substitution

The modifications discussed above affect the formula’s form or structure, but not its constitutive elements. Fundamentally different, therefore, is the other basic type of flexibility: the replacement of an element by another word, a synonym or plesionym. I follow Hainsworth in separating this process from the modifications sensu stricto: “I do not consider the important technique whereby flexibility is obtained by using synonymic words: for a different word means a different formula” (1968:60; see also 1993:5, 13-15). Bloomfield had already done the same with his dichotomy between inflection and verbal variation of repetitions. But Watkins makes no such distinction, accepting the “renewal of one, two, or more members of a formula . . . under semantic identity” as a part of formulaic flexibility (1995:15, cf. 10).

In the following we shall keep substitution by synonyms manageable by insisting that at least two elements of the original formula remain unchanged, as for “the Aśvins mount onto the chariot,” usually:

8.9.8ab ā́ nūnáṃ raghúvartaniṃ / ráthaṃ tiṣṭhātho aśvinā                

but also:

 10.41.2ab prātaryújaṃ nāsatyā́dhi tiṣṭhathaḥ / prātaryā́vāṇam madhuvā́hanaṃ rátham.

However for “to prop apart heaven and earth,” as in:

6.44.24ab ayáṃ dyā́vāpr̥thivī́ ví ṣkabhāyad

the semantically equivalent:

8.41.10de yá skambhéna  ródasī / ajó ná dyā́m ádhārayat

is not similar enough to count as a formulaic variant.

Lexical substitution by non-synonyms leads to the loss of a formula’s identity, that is, its change into a different formula or its dissolution.

Replacing all the elements by allonyms while leaving the syntactic and metric structures unchanged, as in píba sómam(“drink the Soma”) beside jáhi rákṣas (“smite the demon”), leads to “phrase patterns” and “structural formulas” (see above, footnote 8).


12. Formulaic Flexibility and Unity

Unity Preserved: Variants of the Same, “Flexible” FormulaIdentity Lost: A Different Formula
Basic formula:Expanded:Inflected:Substitute a synonym:Substitute an allonym:Substitute two allonyms:
jáhi rákṣasjáhi rákṣas viśváhājaṅghanat rákṣāṃsibhindhi rákṣāṃsipaśya rákṣaspíba sómam
tásya viddhividdhí tásya nastád véda tád jānīhitásya pibavánād éti

13. General Properties of the Candidate Formulas

Although the original search was limited to recurring designator-pairs, subsequent philological examination has shown that well over half of the formulas were actually longer.33

13.1 Length in Words

2. Gr̥tsamada: 2121208
3. Viśvāmitra:123330223
4. Vāmadeva:73010112
5. Atri:5011018
6. Bharadvāja: 62310012
7. Vasiṣṭha:166611232
K Kaṇva:30261942182

13.2 Fixed versus Flexible Formulas

Only one-sixth of the candidate formulas are fully fixed—a far lower proportion than in Homer, where this is thought to hold for half to two-thirds (see above, section 3). Fully fixed formulas range from entire stanzas, such as Book 3’s family-refrain śunáṃ huvema maghávānam

33 It should be noted, with regard to the column in the table entitled “More,” that of the seven formulas with more than six words, five are fixed whole stanzas, and two are three verses in length.


índram / asmín bháre nŕ̥tamaṃ vā́jasātau / śr̥ṇvántam ugrám ūtáye samátsu / ghnántaṃ vr̥trā́ṇi saṃjítaṃ dhánānām (“For blessing we would invoke bounteous Indra, most manly, at this raid, at the winning of the prize of victory, the strong one who listens, (we would invoke) for help in battles, him who smashes obstacles, the winner of prizes”) (occurring first in 3.30.22, repeated fifteen times); over whole verses, like Book 3’s pátyamānas trír ā́ divó vidáthe (“being master at the rite three times a day”); down to the minimal two words, as in Book 5’s ágre áhnām (“at the start of days”). The distinction between fixed and flexible is far from absolute, since some of the flexible formulas are fixed in part, for example, as to two words out of three or as to word order. But five-sixths of the formulas are modified in one way or another, in addition to which comes lexical substitution.

In the R̥gveda much formulaic flexibility arises due to the transfer of formulas between the different meters, leading to reduction, extension, enjambment, new boundaries, and so on. Within Greek epic this could not happen, but it did when Homer was cited in lyric poetry and tragedy.

13.3 Adjectives and Epithets

Among the seventy-seven two-word formulas, the noun-adjective and name-epithet combinations so influential in Parry’s work make up only one-fourth: rayí- suyáma-, Agní- sudītí-, kumārá- Sāhadevyá-, śyená- r̥jipyá-, Agní- dhartár-, rayím rayivánt, vāmá- bhū́ri-, sákhi- pratná-, hotar purvaṇīka, Índravāyū suṣṭutí-, Vāyú- śucipā́-, Agní- rakṣasvín-, Índra- somapā́tama-, girí- párvata-, mártāya ripáve, rā́dhas- áhraya-havyavā́hana- yájiṣṭha-, and hótāraṃ viśvávedasam. All of these are flexible in one way or another except for the fixed vocative hotar purvaṇīka.

13.4 Epithets and Meter

In his pioneering application of Parry’s ideas to the R̥gveda, Gonda rejects the idea that the epithets serve only metric purposes. While admitting that metrics do play a role (1959a:253-57), he finds that the epithets are primarily used to “suit the context” (63 and passim), “conditioned by sense and versification at the same time” (254) to achieve a “harmony between epithet and context” (175) so that “the epithet fits the context perfectly” (66), and that “places are very few where no motive whatever can be discovered for the occurrence of an epithet” (254). Gonda finds the Homeric epithets to be “in wonderful harmony with the situation” (30) as well. The approach seems circular.

The link between a name and its epithet is very loose in the R̥gveda. The two are rarely contiguous, occurring in the same verse only 113 times in the 1,064 occurrences considered below, so that most of these theonym-epithet groups are inherently split and enjambed. Some epithets have distinct preferences as to position within a verse, but the groups of theonym and epithet are so free that they cannot possibly serve any metrical function in the R̥gveda.


Positions of some divine epithets:34

 Verse-initial:Medial:Final:Total:In same verse as theonym:
sūnúḥ sáhasas5390445
jātávedas-13: voc. 11x21: voc. 20x95: voc. 34x12912
draviṇodas-,-dā́-, -dá-91011*309*
ūrjó nápāt-1701181
adrivant-06x, all voc.43, all voc.491
vr̥trahán-195246, voc. 33x11717
hárivant-050x, all voc.2523
vājínīvasu-0020x, all voc.201
divó nápāt-60061
śubhás páti-0120x, voc. 15x210

34 The occasional use of these epithets with other divinities is ignored here; no distinction is made between meters.



*Seven times in a single refrain.

14. Conclusion

The R̥gveda is formulaic oral poetry. The great majority of its formulas is flexible and can be described using Hainsworth’s approach to Homeric modifications. In fact, a higher proportion of the R̥gveda’s formulas is flexible than the Iliad’s; this agrees with its higher overall formulaicity and shows that stylistically, the R̥gveda is in fact more oral in style than Homer—a conclusion strengthened by its higher frequency of unenjambed and coterminous verses and its lower proportion of necessary enjambment (Dunkel 1996:205-07).

Measures of the relative orality of the Iliad and the R̥gveda:

Formulaic overall:1/2 to 2/3 (?)2/3 to 3/4 (?)
Fixed formulas:1/2 to 2/31/6
Flexible formulas:1/3 to 1/25/6
Unenjambed verses:39%57%
Coterminous verses:11%26%
Necessary enjambment:19%12%

15. Beyond Vedic and Greek

The R̥gveda is in fact not the only ancient Indo-European35 text to surpass the Iliad in orality of style. As measured by modes of enjambment, the Roman comedian Plautus (floruit c. 200 BCE) considerably outdoes the Iliad and is very close to the R̥gveda in stylistic orality (Dunkel 1996). This is also true, to a lesser extent, of Terence two generations later.

The relative orality of Plautus, Terence (senarii),36 and Menander (trimeter) as measured by  types of enjambment:

Unenjambed verses:55%48%28%

35 Much Hittite poetry is more formulaic still, as is true of Sumerian and Akkadian.

36 That is, the meter of spoken dialogue; the values are even higher for the long verses spoken as recitative.


Coterminous verses:18%12%4%
Necessary enjambment:15%26%37%

This cannot be ascribed to his main literary model, the Hellenistic comedian Menander (floruit c. 315 BCE), since he is by these measures far more literary in style than any of the texts considered here. Furthermore all three ancient comedians without question used writing to compose and are not formulaic in the least. The reason for the Romans’ oral style of enjambment might have been aural: for success in show business their dialogues had to be readily comprehensible by their public, which was far less literarily sophisticated than Menander’s.

University of Zurich


Aufrecht 1877
Theodor Aufrecht, ed. Die Hymnen des Ṛigveda. 2nd ed. 2 vols. Bonn: Adolph Marcus.

Bergaigne 1878-83
Abel Bergaigne. La religion védique, d’après les hymnes du Rig-Véda. 3 vols. Paris: F. Vieweg.

Bloomfield 1916
Maurice Bloomfield. Rig-Veda Repetitions: The Repeated Verses and Distichs and Stanzas of the Rig-Veda in Systematic Presentation and with Critical Discussion. 2 vols. Harvard Oriental Series, 20 and 24. Cambridge, MA: Harvard University Press.

Bloomfield and Edgerton 1930
Maurice Bloomfield and Franklin Edgerton. Vedic Variants: A Study of the Variant Readings in the Repeated Mantras of the Veda, Volume I: The Verb. Philadelphia: Linguistic Society of America, University of Pennsylvania.

Bronkhorst 2002
Johannes Bronkhorst. “Literacy and Rationality in Ancient India.” Asiatische Studien / Études Asiatique, 56.4:797-831.

Bronkhorst 2016
______. How the Brahmins Won: From Alexander to the Guptas. Handbuch der Orientalistik, 2.30. Leiden: Brill.

Burkert 1995
Walter Burkert. “‘Irrevocabile verbum’: Spuren mündlichen Erzählens in der Odyssee.” In Hören, Sagen, Lesen, Lernen: Bausteine zu einer Geschichte der kommunikativen Kultur: Festschrift für Rudolf Schenda zum 65. Geburtstag. Ed. by Ursula Brunold-Bigler and Hermann Bausinger. Bern: Peter Lang. pp. 147-58. (Also in Burkert 2001:117-26.)


Burkert 2001
______. Kleine Schriften I: Homerica. Ed. by Christoph Riedweg. Hypo-mnemata / Supplement-Reihe, 2. Göttingen: Vandenhoeck & Ruprecht.

Dunkel 1978
George Dunkel. “Preverb Deletion in Indo-European?” Zeitschrift für vergleichende Sprachforschung, 92.1/2:14-26.

Dunkel 1982a
______. “Autour de ā́ r̥g-védique (A Lexical Study).” Indo-Iranian Journal, 24.2:89-102.

Dunkel 1982b
______. “The Original Syntax of Conjunctive *-kwe.” Die Sprache, 28.2:129-43.

Dunkel 1996
______. “The Relative Stylistic Orality of Plautus and Terence, as Measured by Enjambment-Types; With Remarks on Enjambment in Literary Saturnians and Menander.” Die Sprache, 38.2:201-12.

Dunkel 2014
______. Lexikon der indogermanischen Partikeln und Pronominalstämme. 2 vols. Heidelberg: Winter.

Dunkel forthcoming
______. R̥gvedic Family Grammar. Innsbruck: University of Innsbruck.

Erdosy 1995
George Erdosy, ed. The Indo-Aryans of Ancient South Asia: Language, Material Culture and Ethnicity. Indian Philology and South Asian Studies, 1. Berlin: Walter de Gruyter.

Farmer et al. 2004
Steve Farmer, Richard Sproat, and Michael Witzel. “The Collapse of the Indus-Script Thesis: The Myth of a Literate Harappan Civilization.” Electronic Journal of Vedic Studies, 11.2:19-58.

Finkelberg 2004
Margalit Finkelberg. “Oral Theory and the Limits of Formulaic Diction.” Oral Tradition, 19.2:236-52.

Friedrich 2007
Rainer Friedrich. Formular Economy in Homer: The Poetics of the Breaches. Hermes. Einzelschriften, 100. Stuttgart: Franz Steiner.

Gonda 1959
Jan Gonda. Epithets in the Ṛgveda. Disputations Rheno-Trajectinae, 4. ‘s-Gravenhage: Mouton.

Gonda 1959
______. Stylistic Repetition in the Veda. Verhandelingen der Koninklijke Nederlandse Akademie van Wetenschappen. Afd. Letterkunde. Nieuwe reeks, 65.3. Amsterdam: Noord-Hollandsche Uitgevers Maatschappij.

Gonda 1975
______. Vedic Literature (Saṃhitās and Brāhmaṇas). A History of Indian Literature I: Veda and Upanishads, fasc. 1. Wiesbaden: Harrassowitz.


Goody 1985
Jack Goody. “Oral Composition and Oral Transmission: The Case of the Vedas.” In Oralità. Cultura, letteratura, discorsoAtti del Convegno Internazionale Urbino 1980. Ed. by Bruno Gentili and Giuseppe Paioni. Rome: Edizioni dell’Ateneo. pp. 7-17. 

Goody 1987
______. The Interface between the Written and the Oral. Cambridge: Cambridge University Press.

Grassmann 1873
Hermann Grassmann. Wörterbuch zum Rig-Veda. Leipzig: F. A. Brockhaus.

Hainsworth 1962
J. B. Hainsworth. “The Homeric Formula and the Problem of Its Transmission.” Bulletin of the Institute of Classical Studies, 9:57-68.

Hainsworth 1964
______. “Structure and Content in Epic Formulae: The Question of the Unique Expression.” Classical Quarterly,14.2:155-64.

Hainsworth 1968
______. The Flexibility of the Homeric Formula. Oxford: Clarendon.

Hainsworth 1993
______. The Iliad: A Commentary, Vol. III: Books 9-12. Cambridge: Cambridge University Press.

Harris 1957
Zellig S. Harris. “Co-occurrence and Transformation in Linguistic Structure.” Language, 33.3:283-340. (Also in Harris 1970:390-457.)

Harris 1970
______. Papers in Structural and Transformational Linguistics. Formal Linguistics Series, 1. Dordrecht: Reidel.

Heubeck 1988
Alfred Heubeck. “General Introduction.” In A Commentary on Homer’s Odyssey, Volume I: Introduction and Books I-VIII. Ed. by Albert Heubeck, Stephanie West, and J. B. Hainsworth. Oxford: Clarendon. pp. 3-23.

Hoekstra 1965
A. Hoekstra. Homeric Modifications of Formulaic Prototypes: Studies in the Development of Greek Epic Diction. Verhandelingen der Koninklijke Nederlandse Akademie van Wetenschappen, Afd. Letterkunde. Nieuwe reeks, 71.1. Amsterdam: Noord-Hollandsche Uitgevers Maatschappij.

Jamison and Brereton 2014
Stephanie W. Jamison and Joel P. Brereton, trans. The Rigveda: The Earliest Religious Poetry of India. 3 vols. New York: Oxford University Press.

Kiparsky 1976
Paul Kiparsky. “Oral Poetry: Some Linguistic and Typological Considerations.” In Stolz and Shannon 1976:73-106.

Kirk 1962
G. S. Kirk. The Songs of Homer. Cambridge: Cambridge University Press.


Klein 2012
Jared S. Klein. “Categories and Types of Interstanzaic Repetition in the Rigveda.” Historische Sprachforschung, 125.1:191-210.

Lord 1960
Albert B. Lord. The Singer of Tales. Harvard Studies in Comparative Literature, 24. Cambridge, MA: Harvard University Press.

Lubotsky 1997
Alexander Lubotsky. A Ṛgvedic Word Concordance. 2 vols. American Oriental Series, 82-83. New Haven, CT: American Oriental Society. 

Meillet 1923
Antoine Meillet. Les origines indo-européennes des mètres grecs. Paris: Les presses universitaires de France.

Nagy 1976
Gregory Nagy. “Formula and Meter.” In Stolz and Shannon 1976:239-60.

Nagy 1990
______. Greek Mythology and Poetics. Ithaca, NY: Cornell University Press.

Oldenberg 1888
Hermann Oldenberg. Die Hymnen des Ṛigveda, Band I: Metrische und textgeschichtliche Prolegomena. Berlin: Wilhelm Hertz.

A. Parry 1971
Adam Parry. “Introduction.” In M. Parry 1971a:ix-lxii.

M. Parry 1971a
______. The Making of Homeric Verse: The Collected Papers of Milman Parry. Ed. by Adam Parry. Oxford: Clarendon.

M. Parry 1971b [1928]
______. “The Traditional Epithet in Homer.” In M. Parry 1971a:1-190.

M. Parry 1971c [1928]
______. “Homeric Formulae and Homeric Metre.” In M. Parry 1971a:191-239.

M. Parry 1971d [1930]
______. “Studies in the Epic Technique of Oral Verse-Making. I. Homer and Homeric Style.” In M. Parry 1971a:266-324.

M. Parry 1971e [1933]
______. “Whole Formulaic Verses in Greek and Southslavic Heroic Song.” In M. Parry 1971a:376-90.

Poucha 1942
Pavel Poucha. “Schichtung des R̥gveda. Bestimmung des relativen Alters der Lieder des R̥gveda mit Hilfe zahlenmässiger Berechnung. II.” Archív Orientální, 13.3:225-69.

Russo 1966
Joseph A. Russo. “The Structural Formula in Homeric Verse.” Yale Classical Studies, 20:217-40.

Russo 1976
______. “Is ‘Oral’ or ‘Aural’ Composition the Cause of Homer’s Formulaic Style?” In Stolz and Shannon 1976:31-54.


Russo 1997
______. “The Formula.” In A New Companion to Homer. Ed. by Ian Morris and Barry Powell. Mnemosyne, Supplements, 163. Leiden: Brill. pp. 238-60.

Russo 2011
______. “Formula.” In The Homer Encyclopedia. Ed. by Margalit Finkelberg. Chichester, UK: Wiley-Blackwell. pp. 296-98.

Salomon 2003
Richard Salomon. “Writing Systems of the Indo-Aryan Languages.” In The Indo-Aryan Languages. Ed. by George Cardona and Danesh Jain. London: Routledge. pp. 75-114.

Schmidt 1885
Carl Eduard Schmidt. Parallel-Homer; oder, Index aller homerischen Iterati in lexikalischer Anordnung. Göttingen: Vandenhoeck and Ruprecht.

Skjærvø 2012
Prods Oktor Skjærvø. “The Zoroastrian Oral Tradition as Reflected in the Text.” In The Transmission of the Avesta. Ed. by Alberto Cantera. Iranica, 20. Wiesbaden: Harrassowitz. pp. 3-48.

Stolz and Shannon 1976
Benjamin A. Stolz and Richard S. Shannon, eds. Oral Literature and the Formula. Ann Arbor, MI: Center for the Coördination of Ancient and Modern Studies, University of Michigan.

Van Nooten and Holland 1994
Barend A. Van Nooten and Gary B. Holland. Rig Veda: A Metrically Restored Text with an Introduction and Notes. Harvard Oriental Series, 50. Cambridge, MA: Department of Sanskrit and Indian Studies, Harvard University.

Watkins 1976
Calvert Watkins. “Response.” In Stolz and Shannon 1976:107-11.

Watkins 1995
______. How to Kill a Dragon: Aspects of Indo-European Poetics. Oxford: Oxford University Press.

West 2011
M. L. West. The Making of the Iliad: Disquisition and Analytical Commentary. Oxford: Oxford University Press.

Windelberg and Miller 1980
Marjorie Windelberg and D. Gary Miller. “How (Not) to Define the Epic Formula.” Olifant, 8.1:29-50.

Witzel 1992
Michael Witzel. “Alfred Hillebrandt Reconsidered.” Journal of the American Oriental Society, 112.4:611-18.

Witzel 1995a
______. “Early Indian History: Linguistic and Textual Parameters.” In Erdosy 1995:85-125.


Witzel 1995b
______. “R̥gvedic History: Poets, Chieftans and Polities.” In Erdosy 1995:307-52.

Witzel 1997a
______. “The Development of the Vedic Canon and Its Schools: The Social and Political Milieu. (Materials on Vedic Śākhās, 8).” In Inside the Texts, Beyond the Texts: New Approaches to the Study of the Vedas: Proceedings of the International Vedic Workshop, Harvard University, June 1989. Ed. by Michael Witzel. Harvard Oriental Series, Opera Minora, 2. Cambridge, MA: Department of Sanskrit and Indian Studies, Harvard University. pp. 257-345.

Witzel 1997b
______. “Early Sanskritization: Origins and Development of the Kuru State.” In Recht, Staat und Verwaltung im klassischen Indien. Ed. by Bernhard Kölver and Elisabeth Müller-Luckner. München: R. Oldenbourg. pp. 27-52.     

Witzel et al. 2007
Michael Witzel, Toshifumi Gotō, Eijirō Dōyama, and Mislav Ježić, eds. and trans. Rig-Veda: Das heilige Wissen: Erster und zweiter Liederkreis. Frankfurt am Main: Verlag der Weltreligionen.