Ajāmila spends eighty-eight years as a kidnapper, gambler, thief and abandoner of his family. At death he shouts his son's name, which happens to be Nārāyaṇa, with no devotional intent whatsoever. His sins are declared destroyed on the spot, Yama's messengers are sent home, and the god of death subsequently narrows his own jurisdiction so this class of person can't be collected again.
Indra kills a brahmana. He is a god, already devoted, already performing every prescribed rite, and he chants Viṣṇu's name throughout. He is not released. He requires a horse sacrifice, and before that he spends a thousand years hiding inside the fiber of a lotus stalk while the personified sin of brahmin-killing hunts him across the worlds.
Both cases appear in the Sixth Skandha. They are separated by about ten chapters. Neither one acknowledges the other exists.
The first economy is the one Viṣṇu's messengers argue for and Yama ratifies:
"The knowledgeable say that if Vaikuntha's name is chanted, even if it is to address someone else, in jest, in the process of reciting something or negligently, it destroys sins."1
Intent irrelevant. Addressee irrelevant. Mode of speech irrelevant. The mechanism is described as automatic and complete — a phonetic trigger that fires regardless of the state of the person pulling it.
The second economy is what Indra is actually put through, and it runs on entirely different principles: sin as a substance requiring physical distribution and time-based discharge. When Indra kills Viśvarūpa, the brahmin-killing is literally divided into four portions and parcelled out to the earth, trees, women, and water, each accepting a share in exchange for a boon. When he kills Vṛtra, that doesn't work — the sin personifies as a chandala woman and pursues him, and he hides.
"Indra started to suffer from that torment and once it had entered, he could find no peace... In the form of a chandala woman, he saw the sin chasing after him."2
Nothing about that scene suggests a mechanism that name-chanting could short-circuit. It's a chase, with duration, ending in an expensive ritual.
The sharpest evidence that this is a genuine seam rather than two unrelated stories comes from the moment the two economies are quoted in the same breath — and the promise made is the one that isn't kept.
When Indra hesitates to kill Vṛtra specifically because he fears the sin, the rishis reassure him using the first economy's logic, almost verbatim:
"By chanting his name, the slayer of a brahmana, the slayer of a father, the slayer of a cow, the slayer of a mother, the slayer of a preceptor, a sinner, a person who eats dogs, and even a pulkasa, are all purified. With devotion, we must perform a great horse sacrifice. Even if you kill all mobile and immobile objects, along with Brahma, you will not be tainted."3
That's the Ajāmila doctrine, cited to a god, as reassurance, by qualified brahmanas — the same list of grave sinners, the same claim about the name's automatic efficacy. And then Indra kills Vṛtra and is immediately, comprehensively tainted anyway, exactly as he feared.
The promise and its failure sit inside a handful of chapters. The text doesn't note the discrepancy or explain why the assurance didn't hold. Note too that even the reassurance itself is internally mixed: it invokes name-chanting and immediately specifies a horse sacrifice, as though the speakers were themselves unsure which economy was actually operative.
Several reconciliations are available, and the text endorses none of them explicitly, which is worth stating rather than picking one and presenting it as the resolution.
Perhaps the difference is who — Ajāmila is a fallen brahmana at the moment of death, Indra is a god with ongoing cosmic responsibilities, and gods might be held to a standard mortals aren't. But the text's own list explicitly includes killers of brahmanas among those purified by the name, without carving out an exception for divine offenders.
Perhaps it's timing — Ajāmila is dying, with no further actions ahead of him, while Indra continues ruling and acting for ages afterward. A sin destroyed at death has no future conduct to contaminate; a sin destroyed mid-career does. That's coherent, and nothing in the text says it.
Perhaps it's jurisdiction, in the sense the Yama material suggests: name-chanting removes you from Yama's authority over post-death consequence, while Indra's problem is an entirely different kind of contamination operating in life, which Yama never had jurisdiction over anyway. This is the most promising reading, and it's still an inference the text never makes.
Worth looking directly at how the second economy actually discharges sin, because the mechanism is startlingly concrete and has no counterpart in the name-chanting account.
After Indra kills Viśvarūpa, the brahmin-killing doesn't dissolve. It gets divided:
"To purify himself, at the end of a year, Hari divided the elements of this sin into four parts and distributed it among the earth, water, trees and women."4
Each recipient accepts a share and is compensated with a boon. The earth takes a quarter and receives the property that excavated ground refills; the sin remains visible in deserts. Trees take a quarter and gain the ability to regrow severed branches; the sin shows as sap. Water takes a quarter and gains the power to increase whatever it's mixed into; the sin shows as foam. Women take a quarter and receive perpetual desire; the sin shows as menstruation.
⚠ That last clause is the text's own etiology and is reproduced here because the page's argument requires the full mechanism, not because the claim is endorsed. It is stated flatly, without qualification, as part of a cosmological accounting procedure.
What matters structurally: sin here behaves like a conserved quantity. It cannot be destroyed, only moved, and every transfer requires compensating the recipient. That's a fundamentally different physics from "the sun dispels the mist." One economy conserves; the other annihilates.
There's a reason to suspect the distribution mechanism is the substrate and name-chanting is the overlay, though the text itself never says so and this remains inference.
The four-way distribution has the texture of an etiological system — it exists to explain observed features of the world (deserts, sap, foam, menstruation) by tracing each to a cosmic event. That's a characteristic job of older mythological strata: the story explains why things are the way they are. Name-chanting explains nothing about the observable world. It makes a claim purely about soteriology, about what happens to a person, with no accompanying account of any physical phenomenon.
Texts that accumulate over centuries frequently show exactly this layering — an older explanatory stratum concerned with why the world looks like it does, and a later devotional stratum concerned with how a person is saved, coexisting without the later layer ever formally retiring the earlier one. That would account for why both run simultaneously and why nobody in the narrative notices the tension: the two aren't competing answers to one question, they're answers to two different questions that happen to use the same vocabulary of sin and purification.
[n=1 OBSERVED] — this is a single volume's evidence and a plausible reading, not a demonstrated stratigraphy. Testing it would require the same analysis across the other volumes of this text and ideally across other Purāṇas.
You've encountered a system — organizational, legal, religious, familial — that offers two incompatible accounts of how consequence works, and you've noticed that which account applies seems to depend on something nobody will name.
The unhelpful response is deciding one account is the "real" rule and the other is decoration. Usually both are real, both get invoked sincerely by people who believe them, and the switching between them is invisible to the people doing the switching.
The useful move is to stop asking which rule is true and start cataloguing which cases got which treatment. Ajāmila and Indra aren't a contradiction to be resolved by argument — they're two data points, and the pattern that explains them, if one exists, will show up across more cases rather than through closer reading of either one. Note what varied: status, timing, whether the person continued acting afterward, who was making the ruling. That's the actual investigative work, and it requires resisting the pull toward premature reconciliation.
The corollary worth holding: when a system reassures you using one economy and then bills you under the other, the reassurance wasn't necessarily dishonest. The people reassuring you may have genuinely not known which rule was going to apply.
The reassurance and the outcome invoke different mechanisms. Watch specifically for a promise made in one register and a consequence delivered in another — that mismatch is the clearest evidence two systems are running simultaneously.
Even the reassurance is internally mixed. The rishis promise name-chanting and prescribe a horse sacrifice in the same speech. When advocates hedge across both economies without noticing, they likely don't know which applies either.
Nobody in the text remarks on the discrepancy. Genuine seams usually pass unnoticed by the characters. A contradiction that gets acknowledged and argued about is being deliberately staged; one that passes in silence is more likely a real compositional join.
Strongest support: the rishis' reassurance to Indra quotes the Ajāmila doctrine almost verbatim — the same catalogue of grave sinners, the same claim about automatic purification — which means this isn't two unrelated passages that happen to differ. It's the same doctrine, cited in the same volume, failing to deliver in the second case.
Tension, and this page's whole subject. The text runs both economies without reconciliation. This page deliberately does not resolve it, per the standing rule against silently closing contradictions the source leaves open. Every candidate reconciliation above is an inference; none is the text's own position.
Open question. Is there a third case in this volume that would discriminate between the candidate explanations — someone non-divine who chants the name mid-life rather than at death, and either does or doesn't require further atonement? Identifying such a case would test the "timing" hypothesis directly. Not searched exhaustively here.
This vault's Śiva Purāṇa material shows the same corpus family handling internal inconsistency in a comparable way — multiple incompatible accounts of the same event, presented across a volume without any character or narrator flagging the discrepancy. What distinguishes this case is that the inconsistency isn't between two versions of one story, but between two operative rules applied to different stories, both of which are meant to be load-bearing doctrine rather than narrative detail. That makes it a more consequential kind of seam: a reader can hold two incompatible accounts of what happened at a particular sacrifice without it affecting anything they do, but two incompatible accounts of how sin is discharged is a live practical question for anyone taking the text as guidance.
With The Washing of the Elephant (eastern-spirituality). That page documents Parīkṣit's objection that ritual atonement is futile for someone who keeps relapsing, and Śuka's reply — which offers a graduated, sustained-discipline answer (fire reaching the bamboo root) and then, a few sentences later, substitutes an instantaneous one (sun dispersing mist) without acknowledging the switch.
That's the same seam, visible in miniature, one chapter earlier than the Ajāmila and Indra cases make it visible at full scale. What neither page states alone: the two-economy problem isn't a discrepancy between two distant episodes that a careless compiler failed to notice — it's present in compressed form inside a single continuous answer, where the switch happens between adjacent sentences. That substantially changes how to read it. A join between distant passages suggests two source traditions stitched together. A switch inside one speech suggests something closer to a live unresolved question that the text keeps circling and never settles, reproducing the same oscillation at every scale it operates on.
Sharpest implication. When a system reassures you under one rule and bills you under another, the most useful question isn't which rule is genuine. It's what varied between the case you were promised and the case you got — because the switching criterion is usually real, usually consistent, and usually not something anyone inside the system can articulate.
Generative questions.