Who Gets a Speech Tag

The Two Words Before the Quotation Mark

Before anyone in this volume speaks, the text says who is speaking. Śrī-Śuka said. The king asked. Prahlāda said.

It is the most ordinary machinery a narrative has, and it is doing something worth counting. Some figures are named. Some are described by office. Some are only a pronoun. And one being — the one the book is about — is named several hundred times in the narration and never once in a speech tag.

⚠ This page is bounded to Debroy's English. Sanskrit marks reported speech differently from English and the tag inventory below is a property of this translation. Where a feature is demonstrably in the source rather than the rendering, the page says so. That boundary is not a disclaimer at the end; it is part of the finding.

The Census

Every speech tag in the body text of Skandhas 5–10 was extracted by pattern — a capitalised subject, a speech verb, and an opening quotation mark.¹

The count is sensitive to where you set the sentence boundary. A strict boundary — the tag must begin a sentence — returns 308 tagged speech acts across 90 distinct tags. A permissive one, allowing tags mid-paragraph, returns 401 across 101. The figures below that carry weight are identical under both, and the page flags where that is true.

The top of the distribution:

58 of the 90 distinct tags occur exactly once. The speech economy is heavily concentrated at the top — the narrator and his listener — and then almost entirely singletons.

The God Is Never Named in His Own Tag

This is the finding, and it is exact.

Search the volume's speech tags for the god's names and every one returns zero:²

He is referred to by name over five hundred times. He speaks twenty-three times. On none of those twenty-three occasions is he named. He is always The illustrious one.

The three tags reading Vasudeva said are not counterexamples. All three are in Skandha 10 and all three are Kṛṣṇa's father, addressing Kaṁsa and the newborn.²

These counts hold identically under both boundary settings. They do not depend on the extraction choice.

The Rudra Contrast

An absence is only a finding if the text had an alternative. Rudra supplies it.

Rudra speaks four times in this volume. Three of those are tagged Rudra said, and one is tagged The illustrious Rudra said.²

So the volume possesses an epithet-plus-name form, and applies it to the other great deity in the book. It is not the case that divine speech in this text is untaggable by name, or that the honorific crowds the name out. Both can sit together, and for Rudra they do.

What is reserved for one being is the epithet alone — a title with no name attached to it. Viṣṇu is the only figure in the volume who speaks under a tag that identifies an office and nothing else.

The Sanskrit convention behind this is śrī-bhagavān uvāca, which does literally mean "the illustrious one said." So the epithet-tag is genuinely in the text and not Debroy's invention. What remains open, and cannot be settled from a translation, is whether The illustrious Rudra renders a distinct Sanskrit formula or is Debroy disambiguating a bare bhagavān that would otherwise be read as Viṣṇu.

Analytical Case Study: Twelve Voices in One Chapter

The singletons are not scattered evenly, and one chapter accounts for a striking share of them.

At 7(8) — Narasiṃha, immediately after Hiraṇyakaśipu's death — twelve classes of being praise the avatāra in turn, and each gets its own tagged speech: the rishis, the ancestors, the Siddhas, the vidyādharas, the nāgas, the Manus, the Prajāpatis, the gandharvas, the cāraṇas, the yakṣas, the kimpuruṣas, the kinnaras.³

Every one of those twelve tags is a singleton. None of these classes speaks anywhere else in the six Skandhas of this volume.

That is a deliberate structure rather than an accident of a crowded scene. The volume has other crowded scenes — the churning of the ocean, Kṛṣṇa's birth — and they do not do this. At 10(2) the gods praise collectively under a single tag; at 7(8) the same rhetorical job is broken into twelve separately attributed voices.

What the division buys is coverage. A single collective tag says heaven approved. Twelve say every order of being approved, severally, and the enumeration is the argument — it certifies the killing by exhausting the register of witnesses. The chapter needs that certification more than any other in the volume, because what has just happened is a god dismembering a man who held a boon against exactly this.

The scene is a ratification, and the speech tags are the roll-call.

The Pronoun Tags Are the Translator's Layer

Twenty-seven tags in the strict count are bare pronouns: He 19, She 4, They 4.¹

This is where Debroy's hand is most visible. Sanskrit does not need a nominal subject in a speech tag — uvāca carries person and number, and the referent is recoverable from the preceding sentence. Rendering that into English forces a choice: supply the name, supply a pronoun, or restructure. The nineteen He said tags are the places where the translator judged the referent obvious enough not to name.

They are also the places most likely to mislead a reader working at speed, because the referent of He in a text with three or four active male figures per scene is exactly what the reader is least certain about.

This joins a pattern this build has already recorded on the frame narration: four different kings in this volume are tagged The king said, distinguished only by quotation level. The speech-tag layer looks like data and is partly interpretation, and where it is interpretation it is undeclared.

Implementation Workflow

Count tags separately from mentions. The finding here is a ratio — five hundred namings against zero in one position. Neither figure alone says anything. A name census that does not distinguish narration from attribution would show the god as the volume's most-named figure and miss that he is never named where it counts.

Report the boundary setting whenever a tag count is quoted. 308 or 401 depending on where a sentence is judged to begin. Any figure of this kind is a function of the extraction rule, and a count published without its rule cannot be checked or reproduced.

Check whether the load-bearing numbers survive the alternative rule. Here they do, identically, which is the reason the finding can be stated firmly while the totals are hedged.

Treat pronoun tags as translator decisions and flag them. Where the source does not require a nominal subject, every English pronoun tag is a judgement about what the reader can recover. Those are the points where a translation is thinking on the reader's behalf without saying so.

Diagnostic Signs

A figure's mention count and tag count diverge by orders of magnitude. High narration, zero attribution, is a structural fact about how a text lets someone speak.

An epithet-only tag exists alongside an epithet-plus-name tag. The availability of the second form is what proves the first is a choice. Without Rudra, the god's tag would read as the only option the text had.

A cluster of singleton tags in one chapter. Twelve speakers who never speak again, in a single scene, is a rhetorical device rather than a cast list.

Bare pronoun tags in a translated text. These mark where the source omitted the subject and the translator declined to supply it, and they are undeclared interpretive acts.

Evidence, Tensions, Open Questions

Strongest support. All tags were extracted programmatically across the full body text of Skandhas 5–10 under two independent boundary rules, and the god's zero-count was tested separately against five distinct names. The 7(8) cluster was located by chapter and every member verified as a singleton across the whole volume.

Tension the page preserves. Whether the epithet-only tag reflects reverence, formula, or simple inheritance from the genre is not decidable here. The Rudra contrast shows the form is not the text's only option; it does not show what the choice means. The page reports the distribution and declines the interpretation.

Boundedness, restated where it bites. Śrī-bhagavān uvāca is a real Sanskrit convention, so the headline finding is about the text and not only the translation. The pronoun tags and the four-kings ambiguity are the translation's. Whether The illustrious Rudra corresponds to a distinct Sanskrit form is unresolved and is the one place where the finding's sharpest contrast rests on a rendering.

Absence-claim discipline. The god is never named in a speech tag is a claim about a search on five names, each paired with five speech verbs and an opening quotation mark, across the body text of Skandhas 5–10. A tag using a name not on that list — Acyuta, Mukunda, Uttamaśloka, all of which occur in narration — would not have been caught. The list was chosen as the five commonest and is not exhaustive.

Structural finding at n=1 for the chapter cluster. Twelve singletons in one chapter is a single instance of a device, not a pattern. It is offered as a described scene, and the register is held accordingly.

Open question. Does Volume 1 show the same zero? If Kṛṣṇa is ever tagged by name anywhere in Debroy's Bhāgavata, the finding narrows from a property of the text to a property of these six Skandhas — and the Volume 1 ingest running concurrently could answer it with one search.

Author Tensions & Convergences

This page and the frame page are measuring the same layer and reaching the same conclusion from opposite ends.

The frame page found that a tag alone cannot identify a speaker — four kings share one. This page finds that a tag can also refuse to identify a speaker who is perfectly well known. In both cases the attribution machinery carries less information than it appears to, and in both cases the shortfall is systematic rather than sloppy.

Set beside the volume's citation practice, a shape appears. The text has an agentless way of citing an old story, used exactly where the story has to overturn someone with standing. It has an epithet-only way of attributing speech, used exclusively for the highest being in the book. Both are constructions in which an utterance arrives without a person attached to it — one because the transmitter is suppressed, the other because the speaker is de-individuated into an office.

Whether those two are one habit or a coincidence of two conventions is not something this volume will say. What can be said is that the text's most authoritative speech and its most decisive corrections are both delivered by something other than a named individual, and that the volume never remarks on either.

Cross-Domain Handshakes

No load-bearing cross-domain handshake. This page's subject is the speech-tag inventory of one translated volume, and its connections — to the frame narration, to the citation formulae, to the Narasiṃha chapter the roll-call ratifies — are all inside eastern-spirituality and are carried in Connected Concepts.

A handshake could be built here to institutional or de-individuated speech in behavioral-mechanics, and it would be decorative: the vault's page in that area is about authority overriding an actor's conscience, which is a claim about responsibility, not about how a speaker is named. The standing rule is that a connection which does not produce an insight neither domain generates alone should be cut. This one would not.

The Live Edge

Sharpest implication. The being this book exists to describe is named over five hundred times in its narration and zero times in a speech tag. When he speaks, he is only ever the illustrious one — and the volume proves this is a choice by giving Rudra his name and the epithet together.

Generative questions.

1. If the god's utterances arrive attached to an office rather than a person, are they structurally the same object as the volume's agentless citations — utterances with nobody to cross-examine — and does that predict where each is deployed?
2. Twelve classes of being get one speech each, once, to ratify a killing that a boon should have prevented. Is enumerated witness a general device in this corpus for certifying an irregular outcome, and does it recur where other boons are defeated?
3. Nineteen tags in this translation say only He. How many of those are ambiguous in context, and would resolving them change any reading the vault currently holds?

Connected Concepts

• The Frame That Surfaces to Mark a Question — the four kings who share one tag, and the level data that separates them
• The Embedded Correction — the volume's other way of delivering speech with nobody attached
• Narasiṃha, the Liminal Man-Lion — the killing that 7(8)'s twelve voices ratify
• The Asura Boon Loophole — the boon the roll-call is convened to certify around
• The Nested Narrators — the level architecture the tags sit inside
• Debroy's Simplest Reading — the declared method behind the pronoun tags
• Kṛṣṇa Bhakti Hub — the corpus whose central figure is never named in a tag

Footnotes