Instrument · corpus statistics
For each target lemma, every other lemma appearing in the same stanza, scored against what chance would predict. No interpretation is applied. What emerges is a portrait of each figure built entirely from its neighbours — and several of the portraits are not the ones the tradition transmits.
n is the number of stanzas where both words appear. ×exp is how many times more often than chance would predict, given how common each word is in the corpus overall.
A ×exp of 40 means the pairing is forty times stronger than random association. Minimum three co-occurrences; anything below 1.5× is discarded. 164,758 tokens, stanza-level windows.
This is not semantics. It is proximity. Two words can co-occur because one kills the other.
128 stanzas
| Collocate | n | ×exp | Meaning |
|---|---|---|---|
| kṣayádvīra | 4 | 41.2 | "ruling over men" — or "destroying men." The ambiguity is the god |
| bheṣajá | 6 | 17.1 | MEDICINE. His strongest substantive association after his sons |
| mīḍhváṃs | 7 | 14.1 | "bountiful, generous" — an epithet of appeasement |
| babhrú | 4 | 13.7 | Tawny, brown — his colour |
| pṛ́śni | 4 | 11.4 | The dappled cow — mother of the Maruts |
| √mṛḍ | 8 | 9.2 | "be gracious, spare us" |
| ādityá · marút · ásura | 17 · 25 · 6 | 9.9 · 5.1 · 7.0 | His company: the Ādityas, his sons, and the asuras |
Medicine, a colour, a herd-animal mother, a plea to be spared, and a word that means both ruler of men and destroyer of men.
No sacrifice vocabulary. No hotṛ, no yajña, no oblation. Rudra is the only major figure whose collocates are pharmacological and zoological rather than liturgical.
2,305 stanzas — the most frequent figure in the corpus
| Collocate | n | ×exp | Meaning |
|---|---|---|---|
| marútvant | 49 | 3.3 | Accompanied by the Maruts |
| vájra · vajrín | 93 · 54 | 3.1 · 2.9 | The thunderbolt, and the one who holds it |
| śatákratu | 46 | 3.1 | "of a hundred powers" |
| vṛtrá | 123 | 3.0 | The one he kills |
| puruhūtá | 48 | 2.7 | "much-invoked" |
| √han · ójas · śávas · śū́ra · ugrá | — | 2.0–2.3 | Slay, might, power, hero, fierce |
| sóma · máda · índu | 360 · 122 · 126 | 1.7–2.1 | The drink, the intoxication, the drop |
| maghávan · sákhi | 158 · 94 | 2.0 · 1.8 | Bountiful, and FRIEND |
A weapon, a victim, a drink, an intoxication, and a friendship. Every high collocate is either violence or alcohol or patronage.
And sákhi, friend, at 94 co-occurrences. The relationship the poets claim with him is contractual friendship — the same word that governs their relation to a paying king.
| DASYU · 67 stanzas | n | ×exp |
|---|---|---|
| hantár — slayer | 3 | 59.1 |
| avratá — WITHOUT VOWS, lawless | 4 | 57.3 |
| dabhī́ti — a named ally of Indra | 3 | 47.2 |
| ā́rya | 9 | 37.3 |
| vṛ́ka — WOLF | 5 | 26.2 |
| √dhū — shake off · śúṣṇa | 4 · 4 | 16.6 |
| māyā́ — magic, illusion | 5 | 12.5 |
| DĀSA · 25 stanzas | n | ×exp |
|---|---|---|
| námuci — a named individual | 3 | 140.7 |
| śíras / śīrṣán — HEAD | 6 | 56.3 |
| śatá — hundred | 3 | 11.4 |
Dasyu is a category. Its top collocates are ārya at 37×, avrata — "vowless" — at 57×, hantṛ "slayer" at 59×, and vṛka, wolf, at 26×. Defined by opposition to ārya, by having no vows, by being hunted, and by magic.
Dāsa is people with names and heads. Its strongest collocate is Namuci — a named individual — at 141×, and its second is "head" at 56×, from the accounts of beheading. Not ārya. Not avrata.
Dasyu is a slur with a definition. Dāsa is a set of named people who were killed.
Collapsing them into one enemy — which nearly every account does — loses that distinction, and the distinction is visible in the statistics without any interpretation applied.
| Collocate | n | ×exp | Meaning |
|---|---|---|---|
| √muṣ — TO STEAL, ROB | 4 | 70.3 | The action performed on or by them |
| nidhí — HOARD, treasure | 3 | 39.6 | What they have |
| √guh — to hide | 3 | 18.6 | What they do with it |
| √randh — to subdue | 3 | 15.4 | What is done to them |
| revánt — wealthy | 3 | 11.3 | What they are |
| gó — cattle | 10 | 4.0 | The specific property |
Steal, hoard, hide, subdue, wealthy, cattle. Every strong collocate is about property and its transfer. No ārya, no avrata, no colour term.
The Paṇis are rich people who will not hand over cattle. That is the whole portrait, and it is why Bṛbu the Paṇi could be praised the moment he gave thousands away.
| Stanzas | What it keeps company with | |
|---|---|---|
| múni | 5 | Only sákhi, friend, at 26.9×. Too rare for anything else to register |
| keśín — long-haired | 17 | hári (bay horse) 17.0× · √yuj (yoke) 12.4× · √vah (convey) 8.4× · índra |
Keśin is overwhelmingly about Indra's long-maned HORSES, not about ascetics. Fourteen of its seventeen stanzas are chariot vocabulary — yoke, convey, bay steeds.
Only RV 10.136 uses it for the long-haired holy man, and that hymn is in the Popular stratum, the youngest layer. The wild ascetic is one hymn.
So the muni portrait — wind-girt, soiled yellow garments, treading the path of wild beasts, drinking from the cup with Rudra — rests on a single late hymn, and the statistics confirm it has no roots elsewhere in the corpus. That is a real constraint on how much weight the figure can bear.
| Figure | Defining collocates | What it says |
|---|---|---|
| Varuṇa 392 stanzas | mitrá 19.2× · aryamán 20.5× · áditi 9.4× · rā́jan 5.0× · vratá 4.7× · síndhu 4.1× | He never appears alone. Bound to Mitra and the Ādityas, to kingship, to vrata — the vow — and to the Indus |
| Viṣṇu 98 stanzas | vikrámaṇa 107.7× · urukramá 89.7× · śipiviṣṭá 80.8× · padá 9.1× · pūṣán 12.7× | Entirely about STRIDING. Wide-striding, the step, the footprint. No creation, no preservation, no avatars — one action |
| Agni 1,604 stanzas | √idh 5.2× · samídh 5.6× · hótar 3.5× · jātávedas 4.4× · adhvará 2.8× | Kindling, firewood, the priest, the rite. Purely liturgical — the opposite of Rudra's profile |
| Soma 950 stanzas | √pā 5.7× · √su 4.6× · √pū 4.4× · camū́ 7.0× · kaláśa 5.8× · pavítra 4.3× · ádri 3.1× | Drink, press, purify, bowl, jar, filter, stone. An industrial process, described in detail |
Agni's profile is 100% ritual apparatus. Soma's is 100% processing equipment. Rudra's is medicine, colour, a cow, and a plea to be spared.
Two of these gods live inside the sacrifice. The third is standing outside it, and the hymns are negotiating with him.
The stanza is a crude window. Words co-occur in a stanza for metrical reasons as well as semantic ones.
Rare targets give unstable results. Muni at five stanzas cannot support any statistical claim; it is included to show that it cannot.
High ×exp on low n is fragile. Namuci at 140× rests on three co-occurrences. The ranking is suggestive, not decisive.
And proximity is not meaning. Dasyu collocates with hantṛ, slayer, at 59× — that is a relationship of killing, not of similarity. The instrument finds associations. It does not tell you what kind they are.
Sarasvatī returned zero because the lemma form used here did not match the corpus. A known gap, not a finding.