How to Kill a Dragon: Vṛtra and a Sentence Older Than Sanskrit

A sentence that outlived its language

Two words do most of the work in the most famous battle in the Rigveda. They sit at the head of the first verse of hymn 1.32, and a Vedic listener would have heard them coming: áhann áhim, “he slew the serpent.” The verb and its object alliterate, lock together, and resolve a tension the poet has barely had time to raise. Indra has done the deed before the hymn properly begins. Everything that follows is commentary on those two words.

What makes the phrase remarkable is not that it is beautiful, though it is. It is that the same sentence, in the same grammatical shape, surfaces in languages the Vedic poets had never heard of and could not have imagined. A Hittite scribe in Anatolia, writing in cuneiform on clay four centuries before hymn 1.32 was likely composed, told of a storm-god who killed a serpent. A Greek hymn to Apollo describes the god slaying a she-dragon at Delphi. An Old Norse poem sets Thor against the world-serpent. Strip these stories down to their grammatical skeleton, and the same three slots appear every time: a hero, a verb of killing, and a serpent. The verb, in case after case, descends from a single reconstructed Proto-Indo-European root.

This is the thesis the American linguist Calvert Watkins built a 600-page book around in 1995, and it is the spine of this essay. The Vṛtra myth is not only the charter narrative of Indra, the most-invoked god of the Rigveda. It is the best-preserved Indic reflex of an inherited formula that the comparative method can trace back to the language spoken on the Pontic steppe before any of its daughters were written down. To read RV 1.32 carefully is to read a text that is at once intensely local, a Punjab poet praising a rain-bringing warrior, and almost unimaginably old.

The hymn that says it once

The Rigveda refers to Indra’s killing of Vṛtra constantly. It tells the story properly only once. Hymn 1.32, attributed by the Anukramaṇī index to the seer Hiraṇyastūpa Āṅgirasa, is the single sustained narrative; everywhere else the event is an allusion, a stock epithet, a flash of memory the audience is assumed to share.[1] The poem runs to fifteen stanzas in the eleven-syllable triṣṭubh metre, and on linguistic and metrical grounds it reads as a relatively late Rigvedic composition, which is itself worth holding onto: the most detailed version of the oldest story is not the oldest text.[2]

It opens with a flourish of self-announcement.

I will declare the manly deeds of Indra, the first that he achieved, the Thunder-wielder. He slew the dragon, then disclosed the waters, and cleft the channels of the mountain torrents.

(RV 1.32.1, trans. Griffith 1889)

The Sanskrit behind “he slew the dragon, then disclosed the waters” is áhann áhim ánv apás tatarda. The first three syllables carry the whole myth: áhan, third singular aorist of the root han, “to smite, slay,” followed by áhim, the accusative of áhi, “serpent.” Indra strikes, the snake falls, and the waters that the snake had penned up break loose down the mountain. The word translated “dragon” is simply áhi, “snake,” though the creature is vast enough that “dragon” has become the conventional English.[3]

The second stanza supplies the props. The weapon is the vajra, the mace or thunderbolt, and the poet pauses to credit its maker.

He slew the dragon lying on the mountain: his heavenly bolt of thunder Tvaṣṭar fashioned. Like lowing cows in rapid flow descending, the waters glided downward to the ocean.

(RV 1.32.2, trans. Griffith 1889)

Notice the simile in the last line. The released waters run “like lowing cows,” and the image is not decorative. The Vedic imagination repeatedly fuses water and cattle, the two forms of wealth a pastoralist most wants liberated and flowing toward him. The same poetic logic animates the Vala myth, where Indra and the Aṅgirases break open a stone enclosure to free cows that double as dawn-light. Vṛtra holds back the rivers; Vala holds back the herds; in both the heroic act is a release, an unblocking, a restoration of motion to something that was wrongly stopped.[4]

The middle stanzas circle the duel rather than narrating it in sequence. Vṛtra is split like a tree felled by an axe; he is the steer who foolishly challenged the bull; he is, in one of the poem’s sharpest images, a man who cannot hold his drink.

Like a non-warrior who can’t hold his liquor, he provoked the hard-pressing, lees-quaffing super champion. He did not withstand the onslaught of his weapons. He was crushed for having challenged Indra, his features smashed.

(RV 1.32.6, after Jamison and Brereton 2014)

By the end the waters flow, the sun and dawn are produced, and Indra is hailed as king “over what moves and what rests.” The structure is not a line but a spiral: the poem returns again and again to the moment of the blow, examining it from new angles, which is exactly what a community does with a story it already knows by heart.

Aside. The Anukramaṇī’s attribution of the hymn to Hiraṇyastūpa Āṅgirasa is a traditional claim, not a verified fact. These poet-ascriptions are old and often internally consistent (the same name attaches to thematically linked hymns), which is why scholars treat them as plausible working data. But they are a layer of received tradition laid over the text, not the testimony of the composer. The honest position is that we know the clan-style and the period; we do not know the man.

A verb for killing, a noun for a snake

Here the philology begins to earn its keep. The phrase áhann áhim is not just memorable Sanskrit; it is a fossil. The verb han descends from the Proto-Indo-European root reconstructed as *gʷʰen-, “to strike, to slay,” and that root left descendants across the family that no borrowing could explain. Sanskrit han, Avestan jan, Greek theínō and the noun phónos (“slaughter”), Hittite kuen-, Latin -fendō (in dēfendō, offendō), and the Germanic word that becomes Old English gūþ, “battle,” all point back to the same prehistoric verb.[5]

The object word is messier, in a way that is itself instructive. Sanskrit áhi, Avestan aži, and Greek óphis all mean “serpent” and align neatly under a reconstructed *h₁ógʷʰis. A second, partly overlapping snake-word, *h₂engʷʰi- or *h₂n̥gʷʰi-, surfaces in Latin anguis, Lithuanian angìs, Old High German unc, and Greek ékhis. Historical linguists still argue about whether these are one root or two that contaminated each other in prehistory, and the argument matters here because it shows the limit of the method: the verb of killing is securely inherited, the snake is securely a snake, but the exact reconstruction of the noun remains contested.[6]

Watkins’s contribution was to insist that the formula, not the individual words, is the unit of inheritance. He reconstructed the underlying sentence as something close to *(h₁e-)gʷʰent (h₁)ógʷʰim, “(he) slew the serpent,” and argued that what Indo-European poets transmitted was a syntactic template with three obligatory slots: HERO, an act of SLAYING built on *gʷʰen-, and a SERPENT or comparable adversary.[7] The formula could be dressed in different vocabulary in different branches, and the snake could be renamed, but the grammatical shape and the inherited verb kept recurring. The Vedic áhann áhim, where both the verb and the object preserve the inherited roots in their inherited relationship, is the cleanest surviving specimen.

graph TD
    A["PIE: slay the serpent"] --> B["verb root: gwhen-"]
    A --> C["object: serpent"]
    B --> D[Vedic: han]
    B --> E[Avestan: jan]
    B --> F["Greek: theino / phonos"]
    B --> G[Hittite: kuen]
    C --> H[Vedic: ahi]
    C --> I[Avestan: azi]
    C --> J[Greek: ophis]
    D --> K["RV 1.32: ahann ahim"]
    H --> K

The table below lays out the core lexical correspondences. None of these are guesses; they are the standard reflexes given in comparative grammars of Indo-European.

Methods note. A single shared word can be a loan; a borrowed deity name proves contact, not common descent. What the comparative method treats as strong evidence is a shared structure, a formula with the same syntax filled by words that are themselves regular sound-correspondences across the branches. Áhann áhim and the Avestan jan- aži- are not similar because one people heard the story from another. They are similar because both inherited the same sentence from a common ancestor, the way English brother and Sanskrit bhrātṛ are not loans but cousins. The strength of the inference rises with the number of independent branches that preserve the pattern.

What the name Vṛtra hides

Turn the formula around and a second discovery falls out, this one about the snake’s own name. Vṛtra is not, etymologically, a snake. It is a neuter noun built on the root *wer-, “to cover, to enclose, to obstruct,” and it means “the covering,” “the obstacle,” “the blockage.” The serpent is named for what he does, which is to dam the waters, not for what he is. The Rigveda itself preserves this transparency: Vṛtra lies on the mountain holding the rivers shut, and Indra’s act is described less as a killing than as an opening, a vi-vṛ, an un-covering of what was covered.[8]

This is where the most consequential piece of comparative work on the myth comes in. In 1934 Émile Benveniste and Louis Renou published Vṛtra et Vṛθragna, a monograph that set the Vedic vṛtra beside the Avestan divine name Vərəθraγna. The Iranian word is a compound, vərəθra- plus the *gʷʰen- element, and it means “the smashing of resistance,” “the breaking of the obstacle.” Crucially, Benveniste argued that the abstract neuter, “resistance, obstacle,” was primary, and the personified dragon secondary. In the Iranian branch there is no serpent named Vṛtra at all; there is a god, Verethragna, whose name simply means Victory, the overcoming of vərəθra.[9] The Vedic vṛtrahán, “Vṛtra-smasher,” Indra’s most characteristic epithet, is the exact cognate of the Avestan divine name. Two branches inherited the same compound; one turned it into a god of victory, the other into the slayer of a personified obstacle.

The implication reorders the story. If Benveniste and Renou are right, the original Indo-Iranian conception was not “a god killed a snake” but “a god broke a resistance.” The drought-serpent of the Punjab is one vivid dramatization of an abstraction: the obstacle that must be overcome for the world to function, for waters to flow, for cattle to move, for order to win out over inertia. The snake is a metaphor that hardened into a character. Read this way, Vṛtra belongs to the same conceptual family as the cosmic order ṛta (Sanskrit: ऋत, “what is fitted, set in motion”) whose maintenance the Vedic gods exist to guarantee: Vṛtra is ṛta obstructed, and Indra is the force that restores flow.

The same blow struck in Anatolia and at Delphi

The branch that most dramatically confirms the antiquity of the pattern is also the one that diverged earliest. Hittite, recorded in cuneiform from roughly the seventeenth century BCE, preserves a myth read aloud at the purulli festival, a spring rite of renewal. Its subject is the Storm-god’s combat with a serpent named Illuyanka, a word that is itself a compound of two Indo-European snake-roots, the very *h₁ógʷʰi- and *h₂engʷʰi- that gave Sanskrit áhi and Latin anguis. The serpent’s name, in other words, is “snake-snake.”[10]

The Hittite text survives in two versions, and both contain a feature the Vedic hymn shares: the hero does not win cleanly on his own. In one version the serpent first defeats the Storm-god and takes his heart and eyes; the god recovers them by a stratagem involving a marriage and a mortal helper, and only then kills Illuyanka. In the other, the goddess Inara lures the serpent to a feast, the mortal Hupašiya binds it, and the Storm-god strikes. The dragon-slaying, here, requires a confederate.[11]

Watkins read this helper-or-weapon alternation as part of the formula’s deep structure. His summary of the underlying mytheme is almost algebraic: HERO SLAY SERPENT, with the option of WITH A WEAPON or WITH A FRIEND, but characteristically not both at once.[12] Indra has the vajra and the strength-giving Soma; the Hittite Storm-god, weakened, needs Inara and Hupašiya. The Greek material adds the bow of Apollo and the thunderbolt of Zeus, weapons that function as the Greek counterpart of the vajra. The thunder-weapon recurs across the branches with enough regularity that it, too, looks inherited: the dragon-slayer is, very often, a god of the storm.

The Greek evidence is the most diffuse and therefore the most debated. Apollo’s killing of the serpent at Delphi, narrated in the Homeric Hymn to Apollo, uses the aorist éphne, a direct descendant of *gʷʰen-, exactly the verb that stands in áhan. Zeus against Typhon in Hesiod, and Heracles against the many-headed Hydra, supply the multi-headed-serpent and the released-fertility motifs without the inherited verb. Whether all of these are reflexes of one formula or a mix of inheritance and convergence is a live question, and a careful reader should keep the categories separate.[13]

The sin of the dragon-slayer

There is a stanza near the end of RV 1.32 that has puzzled readers for a very long time. After fourteen verses of triumph, the poet asks Indra a strange question.

Whom did you see, Indra, as the avenger of the serpent, that fear entered your heart once you had smashed him, and you crossed the nine and ninety streams in terror, like a hawk through the airy realms?

(RV 1.32.14, after Jamison and Brereton 2014)

The conqueror flees. Having killed the snake, Indra is suddenly afraid of an avenger, and bolts across ninety-nine rivers like a frightened falcon. Nothing in the preceding triumph prepares us for it, and the Rigveda does not explain it. Later tradition rushed to fill the gap: in the Brāhmaṇas and the epics, Indra’s killing of Vṛtra incurs brahmahatyā, the pollution of slaying a brahmin (Vṛtra is reimagined as the son of the priestly artificer Tvaṣṭṛ), and Indra must flee, hide, and be ritually purified.[14] The Mahābhārata elaborates this into a full legal drama, with Vṛtra recast as a devotee and Indra’s victory bought by the self-sacrifice of the sage Dadhīci, whose bones become the vajra.

Here the comparative frame is suggestive rather than conclusive. The motif of the hero who is polluted, exiled, or made to atone for killing the monster is widespread in Indo-European narrative. Heracles is driven mad and must perform his labours in expiation; the Iranian Θraētaona’s victory is hedged with ambiguity; the dragon-fight repeatedly leaves a residue of guilt rather than pure glory. Georges Dumézil built part of his theory of the Indo-European warrior on exactly this pattern, reading Indra’s career as a sequence of “three sins of the warrior,” transgressions that shadow the very function that makes him useful.[15] Whether RV 1.32.14 is an early Indic glimpse of an inherited theme or simply a local puzzle that later priests rationalized, we cannot say for certain. What we can say is that the Vedic poets, at the moment of their hero’s greatest triumph, chose to end not on the kill but on the killer’s fear.

Aside. It is tempting to read the fleeing-Indra stanza as proof of the inherited “guilty slayer” theme, and some scholars do. But the Rigveda gives no motive, names no avenger, and supplies no rite of purification; all of that is later. The disciplined claim is narrow: the stanza shows that even in its oldest sustained telling, the dragon-slaying was not a simple victory, and that ambivalence is something later Indian and other Indo-European traditions also felt. Anything stronger imports the Brāhmaṇas back into the hymn.

What we can and cannot conclude

It is worth being precise about the strength of each claim, because comparative mythology is a field where confident overreach is the besetting sin.

The securest result is linguistic. That áhan and Avestan jan descend from *gʷʰen-, and áhi and aži from a common snake-word, is not speculation; it follows from regular sound laws and is as solid as any reconstruction in the field. That vṛtrahán and vərəθraγna are the same inherited compound is likewise secure, and Benveniste and Renou’s analysis of “obstacle, resistance” as the primary sense has held up for ninety years.

The next claim is strong but not certain: that the recurring sentence “HERO slays SERPENT,” built on the inherited verb and appearing across at least Indic, Iranian, Hittite, and Greek, reflects a genuine Proto-Indo-European poetic formula rather than independent invention. Dragon-fights are admittedly the kind of story humans invent everywhere. What raises the Indo-European case above coincidence is the alignment of the grammar and the vocabulary, not just the plot: the same verb, in the same syntactic role, against a serpent, in branch after branch.[16]

The weakest claims are the interpretive ones, and they should be held most loosely. Whether the myth “is” a drought myth, a New Year cosmogony, a solar allegory, or a ritual charter has been argued every way, and the Rigveda underdetermines the answer. The table below records the main positions without pretending one has won.

The point is not that we are ignorant. We know the words, the formula, and the spread with real confidence. The point is that the further one moves from the grammar toward the meaning, the more the evidence thins, and an honest account marks the transition rather than smoothing over it.

Coda: the oldest thing in the room

The Rigveda is full of things that feel old: the fire-cult, the pressing of Soma, the machinery of oral transmission that carried the text across three millennia without writing. But almost nothing in it is as old as the sentence at the head of 1.32. The hymn itself is late Rigvedic, its poet a near-contemporary of the redactors. The vajra it describes belongs to a metallurgical world the steppe ancestors never knew. Yet inside this late, polished, local poem sits áhann áhim, three syllables that a Hittite scribe and a Greek rhapsode and a Norse skald would each, in their own idiom, have recognized as the start of the same story.

That is the strange gift of the comparative method. It lets us hear, under the Punjab rain and the lowing of the released waters, a sentence that was already ancient when Sanskrit was young. Open RV 1.32 next to the Hittite Illuyanka text and the Homeric Hymn to Apollo, and read the three of them in a row. The plots diverge, the gods have different names, the weapons change. The verb does not. Somebody, somewhere on the grasslands north of the Black Sea, told of a hero who struck a serpent and let the world flow again, and the verb for “struck” is still doing its work, four or five thousand years later, in the mouths of people who have forgotten they ever shared a language.