Everything in this volume is being said out loud by one man to another. Śuka is speaking; Parīkṣit is listening; the king has seven days to live. That situation is the container for all six Skandhas.
Most of the time you cannot see it. The narration runs on for chapters without any reminder that someone is being addressed, and then, occasionally, the king speaks and the room comes back.
The question this page answers is when. It turns out the frame does not surface where the material is hardest or most surprising. It surfaces according to whether anyone else has taken over the telling — and where a second narrator holds the floor, the king goes almost completely silent.
Across the six Skandhas of this volume there are 119 chapters. The outer frame surfaces — Parīkṣit speaks — in 30 of them, asking 33 questions in total.1
| Skandha | Chapters | Chapters where the frame surfaces | Rate |
|---|---|---|---|
| 5 | 26 | 5 | 19% |
| 6 | 19 | 9 | 47% |
| 7 | 15 | 1 | 6% |
| 8 | 24 | 5 | 20% |
| 9 | 24 | 6 | 25% |
| 10 | 11 | 4 | 36% |
| Total | 119 | 30 | 25% |
⚑ Correction to this build's PRD. The PRD gave this as 7 of 119 chapters. The correct figure is 30 of 119. The PRD's number is off by more than a factor of four and is withdrawn; every figure on this page comes from the recount. The 7 appears to have been a stray reference to Skandha 7, which is the outlier the finding actually rests on.
The distribution is not flat, and the two ends of it are far apart: Skandha 6 surfaces the frame in nearly half its chapters, Skandha 7 in one chapter out of fifteen.
Skandha 7 is where the frame effectively stops existing.
Parīkṣit speaks once, at 7(1), asking why an impartial god took sides against the daityas. Śuka begins to answer, and then hands the telling to somebody else — "In this connection, the divine rishi happily recited an ancient account, when he was asked by Ajatashatru at the great sacrifice."2 From that point Nārada is the narrator, answering Yudhiṣṭhira.
And Parīkṣit does not speak again for the rest of the Skandha. Fourteen consecutive chapters, including the whole of the Prahlāda narrative — the volume's longest sustained story and its most doctrinally loaded — pass without the outer frame surfacing once.
Nārada narrates in 11 of the Skandha's 15 chapters.1 The correlation is as tight as this kind of evidence gets at this scale: the Skandha with the most inner narration has the least outer frame.
The mechanism is not mysterious once stated. A frame surfaces when its listener asks a question, and the listener can only ask a question of the person currently speaking. Once Śuka has yielded the floor to Nārada, Parīkṣit's interruptions would have to cut across two levels of narration to reach anyone. So they stop. The questioning role has not disappeared — it has been transferred to Yudhiṣṭhira, who does the asking inside Nārada's frame instead.
This is worth separating from the explanation a reader would reach for first, which is that Skandha 7 is gripping and nobody wants to interrupt. That may also be true and is untestable. The structural account is testable and it predicts the distribution.
Skandha 6 does the reverse, and it does it with no inner narrator holding the floor for any length of time.
Parīkṣit speaks in nine of nineteen chapters, and the questions are of a different character from the volume's norm.
The 33 questions are not one kind of thing, and the split matters for what the frame is doing.
Most of them are requests for more story. I wish to hear about the intelligent royal sage. How did Hari save the lord of elephants? Speak about the reason for the curse. What did Nanda do to warrant such great prosperity?1 These are prompts. They keep the narration moving and they concede everything — the listener wants the next episode and has no quarrel with the last one.
A much smaller group are objections, and they cluster in Skandha 6.
At 6(1) Parīkṣit does not request; he pushes back. Having heard Śuka recommend atonement, he replies that a man who sins, atones, and sins again has achieved nothing — "like the washing of an elephant, I think that atonement serves no purpose."3 That is a listener contradicting his teacher on doctrine, and Śuka's answer restructures the whole account of expiation around devotion.
The Skandha continues in that register: why were the gods abandoned by their own preceptor, how did Vṛtra's mind become fixed on the illustrious one, how did beings born as asuras become devotees of Viṣṇu.1 These are objections about consistency, and they arrive one per chapter because Śuka is answering them personally.
So the two poles have one explanation. Where Śuka holds the floor and is making arguments, the frame is busy; where he has handed the floor to a second narrator telling a story, the frame goes quiet.
A methodological note that changed this page's numbers, and that the vault should carry forward.
Debroy renders narrative level with nested quotation marks: the outer frame in single quotes, material one level in with double quotes, and deeper levels alternating. That typography is the only reliable marker of who is speaking, because the speech tag is not.
Four different figures in this volume are tagged "The king said" or "The king asked": Parīkṣit in the outer frame, and — one level in — Yudhiṣṭhira at 7(10), Citraketu at 6(15), Satyavrata at 8(24), and Purūravas at 9(14).4 A count keyed on the tag alone returns 33 chapters and silently folds inner kings into the frame. It also destroys the finding, because one of the false positives sits in Skandha 7 and doubles its apparent rate.
Keying on the tag plus the quotation character gives 30 chapters and separates the levels cleanly. The correction is small in aggregate and decisive for the one Skandha the argument depends on.
Count the frame, not the impression. Whether a text's outer frame is "present" is not answerable from reading; the sensation of a frame being active is dominated by the most recent chapter. The distribution across 119 chapters is a different object from the memory of it.
Key the count on typography, not on speech tags, in any nested narrative. Where a translator disambiguates levels with quotation marks and not with names, the marks are the data. Verify what a tag means at least once per distinct speaker before trusting a tally built on it.
Ask who holds the floor before asking why a listener is quiet. Silence in a frame narrative usually has a structural cause available before a psychological one. Check whether the listener's counterpart is still the person speaking.
Look for the displaced questioner. Where an outer listener stops asking, the questioning function has often moved rather than vanished. In Skandha 7 it moves to Yudhiṣṭhira, and the volume gains a second interrogator at exactly the point it loses its first.
One Skandha's rate is an extreme outlier in both directions from the mean. 6% and 47% against a 25% average. Extremes at both ends in the same corpus point to a variable, not to noise.
The outlier low coincides with an inner narrator holding most chapters. 11 of 15. The co-occurrence is what converts a distribution into a mechanism.
The frame's one appearance in that Skandha is the handoff chapter. 7(1) is where Parīkṣit asks and Śuka delegates. A frame that surfaces exactly once, to transfer the floor, is doing a specific job rather than being sporadic.
The high-rate section contains the objections rather than the requests. Where a listener is contradicting rather than prompting, expect the frame to be dense — an objection requires the objector's presence in a way a request for narrative does not.
Strongest support. All 119 chapters were segmented programmatically and every occurrence of the outer-frame speech tags was extracted with its following quotation character and its surrounding text, then inspected individually. The four misattributions were caught by that inspection. The Nārada-narration count is a per-chapter presence test across Skandha 7.
Tension the page preserves. Two explanations for Skandha 7's silence are available — structural displacement of the questioner, and the narrative simply being absorbing enough that no interruption is staged. The page argues for the first because it predicts the whole distribution rather than one Skandha, but the second is not excluded and the text offers nothing that would separate them.
Absence-claim discipline. Parīkṣit does not speak again in Skandha 7 after 7(1) is a claim about a search for the tags The king said/asked and Parikshit said/asked followed by an outer-level quotation mark. A frame reassertion phrased differently — a vocative addressed to the king, a narrator's aside — would not have been caught by this method, and vocatives to Parīkṣit certainly do occur inside the narration. The claim is that he asks no questions, not that he is never addressed.
Boundedness to the English. The quotation-mark hierarchy is Debroy's, not the Sanskrit's. Sanskrit marks reported speech with iti and by verb person, not by nested punctuation. The level assignments used here are therefore assertions about this translation's structure, which is what the vault holds.
Open question. The rate for Skandha 10 is 36% and rising against its neighbours, and Skandha 10 continues into Volume 3. If the frame's density tracks the presence of an inner narrator, the rate should stay high wherever Śuka narrates Kṛṣṇa directly and drop wherever the text delegates. That is checkable against the next volume and has not been checked.
The volume's narrators have already been established in this build as stacked and as compromised. What this page adds is that the stack has a measurable cost: each level of delegation removes the level above it from the conversation.
That cost is not neutral. Parīkṣit is the reader's proxy — he is the one who objects, who says atonement looks pointless, who asks how an impartial god can take sides. Where he is displaced, the text loses its built-in sceptic for fourteen chapters, and those fourteen chapters contain the volume's most extreme doctrinal material: a boy who cannot be killed by any means his father can devise, a teaching that openly "differs from the sacred texts of the worlds."
The volume does not appear to notice this, and it does supply a replacement — Yudhiṣṭhira asks good questions. But the replacement is inside the material rather than outside it, which is a different relationship to what is being claimed. An interrogator who is himself a character in the story being told cannot step out of it.
The convergence with the rest of this build: the text's architecture does consistent work that no character or annotation ever comments on. Here the architecture is quotation marks and speech tags, and the work is deciding who is allowed to object.
Silence-as-Status Marker — the diagnostic needs a precondition it does not state.
Lieberman's rule is that speech volume and authority run inverse: the parent who needs no words has authority, the one who escalates has rank without it. The less you have to say to gain cooperation, the more control you have.
Applied to this text the rule inverts and gives the wrong answer. Parīkṣit speaks least in Skandha 7 — once in fifteen chapters — and most in Skandha 6. Read through the status diagnostic, that makes him maximally authoritative in Skandha 7. He is not. His silence there is caused by a third party taking the narrating role, and it coincides exactly with his losing the ability to shape what is said.
The insight neither produces alone: silence reads as status only when speaking was available. The behavioral-mechanics account works from an implicit assumption that the quiet party could have spoken at any moment and chose not to — which is what makes the silence a signal. Where the floor has structurally moved elsewhere, the identical observable, a person saying nothing, means displacement instead. The rule therefore requires a floor-availability check before it can be applied, and that check is not in the framework because in the parent-child cases it is always trivially satisfied.
Sharpest implication. Across 119 chapters this volume's listener asks 33 questions in 30 chapters. In Skandha 7 he asks exactly one, in the chapter where the telling is handed to Nārada, and then says nothing for fourteen chapters — the longest and most doctrinally extreme stretch in the book runs with the reader's proxy removed from the room.
Generative questions.