Fifty-Seven Hymns, One Last Line: How the Rigveda's Families Signed Their Work

The Line That Will Not Stop Arriving

Read the seventh book of the Rigveda straight through, in any translation, and somewhere around the fifteenth hymn you begin to distrust your own eyes. A hymn to Agni ends: do ye preserve us evermore with blessings. Then a hymn to Indra ends the same way. Then Dawn, then the Waters, then Varuṇa, then the Maruts, then a hymn about rain, then a hymn about the wind. The deities change, the imagery changes, the poet’s argument changes, and the last line does not move.

In Sanskrit it runs yūyaṃ pāta svastibhiḥ sadā naḥ: “do you protect us always with your blessings.” Five words, an imperative addressed to a plural “you,” and the only genuinely fixed thing in a book of 104 hymns. It closes 57 of them. Nothing else in the Rigveda behaves quite like this, and outside the technical literature it is rarely remarked on, partly because the English translation most readers meet renders the line a little differently almost every time it appears, so the effect dissolves into ordinary piety.

It is not piety. It is a mark, and it belongs to a family.

Book 7 is the Vasiṣṭha book, one of six so-called family collections (Books 2 through 7) that form the oldest identifiable stratum of the Rigveda, each associated by tradition with a single clan of poets. [1] [6] What the repeated line suggests is that these clans did not merely inherit a stock of hymns. They had a house style with a visible edge to it: a closing formula that said, to anyone who knew the repertoire, this one is ours. Seven other lineages turn out to have had lines of their own. And because a formula is a physical piece of text rather than an opinion, those lines can be chased through the rest of the collection, into books where they do not belong, and used to ask a question that sounds impossible for a text with no manuscripts, no colophons and no authors’ names attached to anything: can we check who wrote what?

Methods note. Every count here comes from my own pattern search over a romanised text of the Saṃhitā, hymn by hymn, matching the formula and then reading the hits by hand. Three cautions. First, my source text is unaccented, so I quote Sanskrit without the Vedic pitch marks rather than invent them. Second, Vedic is transmitted in sandhi, with words fused across boundaries, so a formula can hide from a naive search; treat my figures as floors. Third, and more amusing, the romanised Saṃhitā text in widest online circulation does not set repeated lines out again; it abbreviates them to an ellipsis, a space-saving convention inherited from print. About 470 of its verses carry an ellipsis in place of words, and 49 of those stand for this one Vasiṣṭha line. Whoever prepared the text got tired of it well before verse fifty-seven. I expanded the abbreviations against the full forms before counting.

The Same Quarter-Line, Fifty-Seven Times

Take a representative case. RV 7.7 is a short Agni hymn, seven verses, and its last verse does two things at once:

nū tvām agna īmahe vasiṣṭhā īśānaṃ sūno sahaso vasūnām

iṣaṃ stotṛbhyo maghavadbhya ānaḍ yūyaṃ pāta svastibhiḥ sadā naḥ

“Now we, the Vasiṣṭhas, entreat you, Agni, master of goods, son of strength. You have brought food to the singers and to the patrons. Do you protect us always with your blessings.”

Rigveda 7.7, verse 7. My rendering from the Saṃhitā, checked against Griffith (1896), who has “We, the Vasisthas, now implore thee, Agni.” [8]

The family names itself and then signs off, in the same breath. That pairing recurs: the word vasiṣṭha appears in 41 verses of Maṇḍala 7 and in only two verses anywhere else in Books 1 through 9, plus eight in the late Maṇḍala 10, which is already a strong statement about who was curating this collection. [1]

The formula is a single pāda, and it sits in the closing quarter of a stanza in triṣṭubh, the eleven-syllable line that is the workhorse metre of the family books. That has a consequence the poets could not argue with. A triṣṭubh quarter-line will not bolt onto the end of a hymn in gāyatrī, whose lines run to eight. And sure enough, of the 57 Maṇḍala 7 hymns that carry the sign-off, 53 are in triṣṭubh and one more in the twelve-syllable jagatī. Of the 47 that lack it, 14 are in short metres where the line simply would not fit.

Which leaves 33 hymns in the right metre that do not take the line. Metre is a gate, not an explanation. Something else decided, hymn by hymn, whether the mark went on, and the text does not say what. The honest position is that we can see the constraint and not the choice.

Aside. Three of the refrain-bearing hymns are in shorter or irregular metres, and one of them is worth the detour. RV 7.34 runs for 25 stanzas of two short lines each, roughly half the bulk of the triṣṭubh stanzas on either side of it, and it still ends with the family line. Whoever put it there wanted the mark badly enough to hang it on a stanza shaped wrong for it.

Every Family Had One

Once you know what you are looking for, the other books give themselves up quickly. Sort the 1,028 hymns by their closing words and the recurrences cluster, book by book, exactly where the tradition says one clan was in charge.

The book column names the home collection and the count includes only the hymns it closes there; the handful of strays elsewhere are the subject of the next two sections. One hymn, RV 2.11, manages to stack two of Gṛtsamada’s formulas into a single closing stanza, which is why 125 formula-endings fall on 124 hymns.

Two things in that table deserve a second look. The first is that the formulas are not interchangeable prayers. They are characteristic. Bharadvāja’s line asks for a hundred winters with good sons, a wish this blog has met before in the Rigveda’s long argument against dying. Gṛtsamada’s asks for something stranger and more social:

asmāñ ca tāṃś ca pra hi neṣi vasya ā bṛhad vadema vidathe suvīrāḥ

“Lead both us and them onward to what is better. Loudly may we speak at the rite, rich in heroes.”

Rigveda 2.1, verse 16. My rendering; Griffith has “With brave men in the assembly may we speak aloud.” [8]

Speaking loudly in the vidatha, the ritual gathering where goods and words were distributed, is a request for standing. It is the wish of a professional who wants his voice to carry in the room, and it is the last thing five of Gṛtsamada’s hymns say. Viśvāmitra’s clan, by contrast, ends on a battlefield, calling Indra “best of men where the prize is won.” Atri’s family asks the Aśvins for cargo. These are house priorities, set down in the one position in a hymn that nobody forgets.

The second thing is the Kutsa line, second from the bottom. Its three Maṇḍala 1 hymns have a fourth relative sitting in a book where it has no business being. Hold on to that. It becomes the key to the best case in this whole investigation.

The Book With No Sign-Off

Maṇḍala 8 is the awkward one. It has 103 hymns, it is traditionally the book of the Kāṇvas, and by the logic above it should be thick with a Kāṇva line. It has none. Exactly one hymn in the whole book closes with a formula that any other hymn also uses, and that one is not a Kāṇva formula at all.

Part of the answer is metre, and it is quantitative. Sort every hymn by the typical length of its stanzas and the books separate cleanly:

Five of the six family books are dominated by eleven- and twelve-syllable stanzas. Maṇḍala 8 is dominated by the short strophic metres, the compact gāyatrī and bṛhatī structures built in pairs and triplets, which is one of the clearest formal facts about it. [13] A book composed almost entirely in eight-syllable lines has no landing site for a triṣṭubh sign-off. The Kāṇvas were working in a different shape, and the mark the triṣṭubh families used would not fit in it.

That is satisfying but incomplete, because it does not explain the single exception, and the exception is the most quietly spectacular thing in this article. RV 8.80 is a ten-verse Indra hymn, and it ends:

tasmā u rādhaḥ kṛṇuta praśastam prātar makṣū dhiyāvasur jagamyāt

“For him make your bounty renowned. May the one rich through insight come early and soon.”

Rigveda 8.80, verse 10. My rendering from the Saṃhitā. The same line closes six hymns in Maṇḍala 1, RV 1.58 and RV 1.60 through RV 1.64.

Those Maṇḍala 1 hymns are traditionally the work of Nodhas Gautama. And the poet to whom the index assigns RV 8.80 is named Ekadyū Naudhasa: Ekadyū the descendant of Nodhas. [4] In the one hymn of the Kāṇva book that carries an inherited sign-off, the sign-off and the poet’s patronymic point at the same family. Nobody arranged that for our benefit. It is what a working habit looks like from three thousand years away.

Could This Just Be Repetition?

A sceptic’s objection arrives here, and it is a good one. The Rigveda repeats itself constantly. Maurice Bloomfield spent two Harvard volumes cataloguing the repeated verses, distichs and stanzas of the text and reckoned that something on the order of a fifth of it is repeated material in one form or another. [2] Jan Gonda later devoted a four-hundred-page study to repetition as a Vedic stylistic principle rather than an accident. [3] In a corpus like that, some lines will recur in clumps for no reason worth a paragraph.

The objection is answerable, because clumping by chance is a thing you can put a number on. Suppose the 57 hymns that end with the Vasiṣṭha line were sprinkled at random across all 1,028 hymns of the Rigveda, with no regard to who composed them. The chance that every one of the 57 lands inside the 104 hymns of Maṇḍala 7 is the hypergeometric tail

$$P = \frac{\binom{104}{57}}{\binom{1028}{57}} \approx 10^{-64}$$

which is not a number that needs interpreting. Whatever put that line at the bottom of those hymns knew which book it was working in.

Aside. The null model is deliberately crude and it flatters the conclusion, since hymns are not independent draws and the metres are not evenly spread. Tighten it by conditioning on metre, restricting the pool to the roughly 660 hymns in the long metres, and the exponent moves by a few places. It does not move by sixty. The point of the calculation is not precision but scale: no reasonable adjustment brings a coincidence into range.

The Seam in the Soma Book

Now the formulas start to earn their keep, because a fingerprint is only interesting once it turns up somewhere it should not.

Maṇḍala 9 is the Rigveda’s structural oddity. Its 114 hymns are addressed almost entirely to Soma Pavamāna, the drink being pressed and filtered, and unlike the family books it was assembled by subject rather than by clan: the redactors pulled the soma material together into one place. [6] [1] The standard reading is that these hymns were not composed for a separate collection but extracted, and that the family repertoires are where they came from. The best support for the idea used to be circumstantial. The sign-offs make it visible.

Nine hymns in Maṇḍala 9 and four in Maṇḍala 10 end with a formula that belongs to one of the family books. If the extraction story is right, each of those should have been written by someone from the family whose line it carries. That prediction is testable, because a much later Sanskrit index, Kātyāyana’s Sarvānukramaṇī, records a poet’s name for every hymn in the Rigveda, compiled from a tradition of authorship that had nothing to do with counting refrains. [4] [5]

Line the two up.

Nine clear agreements, one probable, one refusal. For a test run on a text with no title pages, that is a strong result, and the shape of it matters more than the score: the agreements are not vague family resemblances but exact hits on patronymics, Naudhasa beside Nodhas, Vāsiṣṭha beside Vasiṣṭha, Vaiśvāmitra beside Viśvāmitra.

The longest hymn in the Rigveda has seams you can see

Look again at the three separate rows for RV 9.97, and then look at what that hymn is. At 58 verses it is the longest single hymn in the collection, built out of tṛcas, three-verse units, with one extra stanza at the end. [7] The Anukramaṇī does not assign it to one poet. It walks down the hymn handing out tṛcas: verses 1 to 3 to Vasiṣṭha, 4 to 6 to Indrapramati Vāsiṣṭha, 7 to 9 to Vṛṣagaṇa Vāsiṣṭha, 10 to 12 to Manyu Vāsiṣṭha, and so on through eleven poets of that clan, before handing the last fourteen verses to Kutsa Āṅgirasa. Read as literature this looks like a scholiast’s fantasy, a committee hymn invented to tidy up a monster.

Then you check the closing lines of the units.

abhi svara dhanvā pūyamāno yūyaṃ pāta svastibhiḥ sadā naḥ

“Roar out, run forward, as you are purified. Do you protect us always with your blessings.”

Rigveda 9.97, verse 3, the end of the first tṛca. My rendering; Griffith, “Do ye preserve us evermore with blessings.” [8]

The Vasiṣṭha sign-off closes verse 3 and again verse 6: the exact boundaries of the two tṛcas the index assigns to Vasiṣṭha and to a Vasiṣṭha son. And then, 52 verses later, the hymn ends on a line that has no business in the Soma book at all:

tvayā vayaṃ pavamānena soma bhare kṛtaṃ vi cinuyāma śaśvat

tan no mitro varuṇo māmahantām aditiḥ sindhuḥ pṛthivī uta dyauḥ

“With you as the purifying one, Soma, may we always gather up the stake in the contest. May Mitra, Varuṇa, Aditi, Sindhu, Earth and Heaven grant us this.”

Rigveda 9.97, verse 58. My rendering, checked against Griffith. [8] The second line is word for word the closing line of RV 1.94, RV 1.95 and RV 1.96.

Those three Maṇḍala 1 hymns are the Kutsa hymns. The index gives RV 9.97’s last fourteen verses to Kutsa Āṅgirasa, and the hymn signs off with Kutsa’s line, carried across from another book. Two systems built for entirely different purposes, one a poet’s professional habit and the other a catalogue compiled centuries later by someone counting metres and cataloguing deities, agree on where the joins are inside the Rigveda’s longest hymn.

graph LR
    A["Vasistha repertoire, Book 7"] -->|"sign-off travels with verses"| B["Soma material, Book 9"]
    A -->|"same line"| C["Late hymns, Book 10"]
    G["Kutsa repertoire, Book 1"] -->|"same mechanism"| B
    B --> D["RV 9.90, RV 9.97 vv. 1-6"]
    B --> H["RV 9.97 vv. 45-58"]
    C --> F["RV 10.122"]
    C --> K["RV 10.65"]
    D --> I["Index names a Vasistha"]
    F --> I
    H --> J["Index names Kutsa"]
    K --> L["Index names Vasukra: mismatch"]

Aside. One caution has to be put on the table, because it is the objection a careful reader will raise. Could the tradition have derived its attributions from the refrains, making the agreement circular? For some hymns the tradition plainly did reason from internal evidence, and where a hymn names a poet the index tends to name him too. But the Sarvānukramaṇī is a terse catalogue of poet, deity and metre, not an argument, and its compilers had access to material we have lost; what it does not do is show its work. [4] The circularity cannot be excluded in principle. What can be said is that the agreement holds at the level of tṛca boundaries inside a 58-verse hymn, which is a finer grain than a rule of thumb about refrains would give you.

When the Signature Disagrees

The one clear failure is instructive, so it should not be buried. RV 10.65 is a hymn to the All Gods, and the index assigns it to Vasukarṇa Vāsukra, a poet with no Vasiṣṭha connection. Its last verse:

devān vasiṣṭho amṛtān vavande ye viśvā bhuvanābhipratasthuḥ

te no rāsantām urugāyam adya yūyaṃ pāta svastibhiḥ sadā naḥ

“Vasiṣṭha has praised the immortal gods, who stand over all creatures. Let them grant us wide-ranging space today. Do you protect us always with your blessings.”

Rigveda 10.65, verse 15. My rendering; Griffith, “Immortal Gods have I, Vasiṣṭha, lauded.” [8]

The verse carries the Vasiṣṭha sign-off and names Vasiṣṭha as the speaker, and the tradition still puts someone else’s name on the hymn. Two readings are available. Either a later poet put on a borrowed voice, quoting the grand old family formula and the grand old name the way a modern writer might sign a pastiche, or the index is simply wrong here. Neither is embarrassing. The interesting part is that the two kinds of evidence can disagree at all, which is what makes them evidence rather than decoration. The same logic runs through a century and a half of work on how the collection was sorted into the shape we have.

What a Last Line Is For

Nothing in the Rigveda explains the practice, so the functions have to be reasoned out from what the formulas do.

The plainest one is procedural. Vedic ritual is a coordinated performance in which several specialists take turns, and the priest who recites is not the priest who acts next; the division of labour is explicit in the text itself. A hymn is not a written object with a visible bottom edge. It is sound, and somebody has to know when it has stopped. A fixed closing line is the cleanest possible solution: not a pause, not a change of tone, but a specific string of words that everyone present recognises as the end. Modern liturgies do the same thing with “world without end, amen.” The Rigvedic version is drier and more transactional, which is the house manner throughout: the last thing the Vasiṣṭhas say before falling silent is a request for protection, and the last thing the Gṛtsamadas say is a request to be heard.

The second function is the one this article has been leaning on: ownership. Not authorship in the modern sense, since these poets sign as families rather than as individuals, and not a colophon, which is a scribal habit belonging to a much later world of manuscripts. Something more like a maker’s mark on a tool. It travels with the object, it identifies the workshop rather than the workman, and it is legible only to people who already know the trade. Poets who described their own composing as carpentry were not likely to find a workshop mark strange.

The third is not a function at all but a side effect, and it is the reason philologists care. A mark that means “this ends here” becomes a record of editing wherever the text keeps going. RV 7.1 is the case to look at. It is the longest Agni hymn in the Vasiṣṭha book, and in the text as transmitted its sign-off stanza appears twice: once as verse 20, then again, word for word, as verse 25. Griffith translates the two stanzas identically, as he has to. [8] Read that straight and you are looking at the join: a hymn that closed at twenty verses, four further verses added by someone who wanted them in, and the closing stanza copied out again underneath to shut the thing properly a second time. Hermann Oldenberg built an entire method on inferences of exactly this kind, using metre, refrain and formula to prise apart what had been welded together, and his Prolegomena of 1888 remains the foundation of the subject. [7] [18] Bloomfield’s two volumes are, among other things, the raw index such work needs. [2]

Reading for the Last Line

There is a particular pleasure in a text that turns out to contain its own metadata. The Rigveda has no author list, no dates, no introduction, no editor’s preface. What it has instead is habit, deposited in the words themselves by people who were not trying to leave us anything and could not have imagined a reader without a teacher. The families closed their hymns the way they always had. Three thousand years later those closings are still doing work: separating one clan’s repertoire from another’s, marking where the soma hymns were cut out of the family books and glued into a topical anthology, and exposing the point in the longest hymn of the collection where one poet stopped and another started.

It also changes what the 1,028 hymns feel like. The Rigveda is easy to read as an undifferentiated sacred mass, and the sign-offs make that impossible. They put professionals back into the picture: competitive lineages with house styles, interested in standing and payment and being heard in the assembly, marking their work so that it stayed theirs. The formulas are the least mystical thing in the text and among the most human.

If you want to see it for yourself, do not read Maṇḍala 7 forwards. Open a translation at the seventh book, read only the final verse of each hymn in order, and watch the line arrive, and arrive, and arrive. Then jump to RV 9.97 and read verses 1 to 6 and verse 58, and nothing else. You will be looking at three poets’ handwriting on one page, in a text that never learned to write.