There is a grammarian's example the compendium reaches for when it wants to make a point about reading, and it is a good one because it is annoying. Under a certain technical rule, a dry channel where water used to run counts as a river, and the Ganges and the Yamuna — actual rivers, full of actual water — do not. Anyone who has not studied the rule will hear this as obviously false. Anyone who has studied it will hear it as obviously true, and will also know that the rule was never trying to tell you which things are wet.
The trap is not that the statement is false. The trap is that it is answering a question you were not asking, in a register you did not notice you had left. You brought a question about water. The sentence was about classification. Both parties are being perfectly precise and the misunderstanding is total.
Almost every argument about whether a sacred text is literally true has this shape. Someone brings a question about water to a sentence that was never about water — and the tradition that produced the sentence often has a name for the register it is written in, and a formal account of how to read it.
Pūrva Mīmāṃsā — the ritual-exegetical school whose entire business is determining what Vedic sentences oblige — divides scriptural statements by function rather than by content. Some sentences are vidhi, injunctions: do this. Some are mantra, to be recited. And a large body of material is arthavāda: explanatory, laudatory, or narrative statements that surround an injunction and give it force.1
The school's position on arthavāda is what makes it useful here. These statements are not decorative, and they are not lies. They are subordinate — their function is to motivate and situate the injunction, and their truth conditions are set by that function rather than by independent reference. A passage praising a ritual by describing extravagant consequences is doing its job when it produces the disposition to perform the ritual. Reading it as a report about consequences is a category error of exactly the riverbed kind.
The rendering names three divisions, and the distinctions matter for how far the device can be pushed:1
Guṇavāda — statements whose surface sense is contradicted by other means of knowledge, and which are therefore to be taken figuratively. The standard shape is a metaphor stated as an identity.
Anuvāda — statements restating what is already established elsewhere. They add emphasis rather than information, and are not independent evidence for anything.
Bhūtārthavāda — statements whose surface sense is not contradicted, and which may be taken as they stand.
The third category is the one that keeps the framework honest, and it is routinely forgotten by people who invoke arthavāda casually. The default is not "read figuratively." The default is to read plainly unless something else forces otherwise, and the forcing has to come from an independent means of knowledge, not from finding the plain sense distasteful or implausible by present standards.
Pt. Dinanath Sharma Shastri applies the machinery to a passage that has embarrassed modern readers of the Śiva Purāṇa: Śiva's seed, disturbed at the sight of Viṣṇu's Mohinī form, carried by the Saptaṛṣis and conveyed through Añjanā's ear into her womb.2
His move is not to deny the passage or allegorise it into vapour. It is to identify the register. A god's retas is not physical matter; it is tejas, a radiance. Once that identification is made, the narrative sequence stands as written — the disturbance, the carrying, the ear — while the substance being transmitted is understood as something other than biological. The passage is guṇavāda in respect of the substance and plain narrative in respect of the events.
This is a more disciplined move than it first appears. It does not let the reader relocate whichever details they find awkward. It changes exactly one term and leaves the rest under normal reading, and it justifies the change by a doctrine about divine nature held independently of this passage — which is precisely the "contradicted by another means of knowledge" condition that guṇavāda requires.
The sharpest instance is the essay's handling of Vāyu's approach to Añjanā, because here the interpretive problem is not metaphysical delicacy but a live ethical charge: a god fathering a child on a married woman.
The essay's answer is to read the union as mental and non-physical, and it grounds the reading in the text itself rather than in the interpreter's discomfort. Vālmīki's own narration has Vāyu tell Añjanā that the son will be strong, wise, immensely radiant, and Vāyu's equal in leaping — and, in the same breath, reassure her explicitly that her chastity remains untouched.3 The exculpation is not being supplied by the commentator; it is in the source, spoken by the figure whose conduct is at issue.
The essay then reinforces this from outside the Rāmāyaṇa. The Mahābhārata's Āśramavāsika-parva enumerates five modes by which deities may father children — through will, speech, sight, touch, or contact — none of which compromise human chastity.4 The point of the citation is that a general doctrine exists, is stated as doctrine rather than narrative, and is not being invented for this case.
What makes this a good case study is that it shows the framework's actual test. A bad arthavāda argument reads figuratively wherever the plain sense is inconvenient. This one reads figuratively where an independent doctrine says the plain sense cannot hold, and cites both the doctrine and the text's own self-description. Whether one accepts the conclusion, the procedure is checkable — which is the only property that distinguishes exegesis from motivated reading.
The obvious hazard is that arthavāda can be made to absorb anything. Any passage that costs you may be declared laudatory-subordinate and thereby neutralised, and the school's own literature is aware of this, which is why the three-way division exists at all.
Two constraints do the limiting work. The first is that guṇavāda requires external contradiction — another valid means of knowledge must rule the surface sense out. Finding a passage repugnant is not such a means. The second is that arthavāda is defined by its relation to an injunction; a passage is subordinate to something. Free-floating narrative with no ritual injunction in view is not straightforwardly arthavāda at all, and applying the device to Purāṇic story-material — as the essay does — is already an extension beyond the school's home territory, made without argument.
That extension is worth flagging rather than waving through. Pūrva Mīmāṃsā built this apparatus for Vedic ritual sentences. Its transfer to Purāṇic narrative is a real move with real consequences, and the essay performs it silently.
You are reading a passage that says something impossible, and you feel the reflex to decide: true or made up. Hold the reflex and ask a different question first — what is this sentence for? Is it obliging you to do something, restating a claim made elsewhere, or working to make you feel the weight of something you are already meant to do? The answer changes what would count as the sentence being wrong. A sentence whose job is to motivate is not falsified by the failure of its imagery to describe.
Now you are the one about to read a line figuratively, and you notice the reason is that its plain sense embarrasses you in front of an audience you have in mind. That is not a licence. The test is whether something you hold on independent grounds actually rules the plain reading out — and you have to be able to name that something, out loud, before you touch the passage. If the only thing ruling it out is that the plain reading is unfashionable, the plain reading stands and you write around your discomfort instead.
And here is the harder discipline: having decided a term is figurative, you change that term only. The temptation is to let the figurative reading spread outward until the whole passage becomes conveniently weightless. Watch the boundary. One word moved, the rest under normal reading, and the boundary itself stated where the reader can see it — because a reader who cannot see where your figurative reading stops cannot check it, and a reading nobody can check is not an argument.
The strongest support for the essay's use of the device is that it does not deploy arthavāda opportunistically. It reads the Vāyu episode figuratively on stated doctrinal grounds and reads the Śatarudra Saṃhitā's epithets plainly — including epithets that a modern reader would find easier to allegorise. The selectivity has a rule behind it, and the rule is announced.
The tension the essay does not close. The compendium simultaneously prohibits selective reading of a conceded source — see Ardha-Kukkuṭī-Nyāya — and licenses a technique for reading selected parts of a source non-literally. Both positions appear in the same volume, aimed at different opponents. The half-hen essay attacks skeptics for dividing the Rāmāyaṇa; this essay divides the Śiva Purāṇa. The difference the compendium would presumably assert is that arthavāda is a principled division with a stated criterion while the skeptic's is not — but that argument is never made, and the reader has to construct it.
A second tension: the transfer from Vedic injunctive context to Purāṇic narrative is performed without justification, and it is exactly the move on which the whole application depends.
Open questions. Does arthavāda apply to narrative that is subordinate to no injunction, and if so what constrains it? Would the essay accept a critic using guṇavāda against a passage the compendium wants read plainly — the Rudrāṃśa epithets, say — and on what grounds would it refuse?
Dinanath Sharma Shastri stands in direct opposition to Literal Reading of Six Kaula Terms, which argues that a tradition's plain reading is frequently the historically accurate one and that figurative re-readings are typically later sanitising. That page and this one describe the same operation with opposite valuations: what one calls the recovery of intent, the other calls the loss of the actual practice. Both cannot be generally right, and the vault should hold them against each other rather than let the more recent page settle it.
He converges with Direct Meaning Hermeneutics more than the surface suggests: bhūtārthavāda is a direct-meaning default, and the essay's whole architecture assumes plain reading holds unless displaced.
To Ta'wīl: The Return Hermeneutic. Both are formal, tradition-internal licences for non-literal reading, and both exist to prevent an embarrassing passage from either falsifying a revelation or being quietly ignored. The structural difference is the direction of authority. Ta'wīl returns a passage to its origin — the deeper sense was always there, and the interpreter ascends to it. Arthavāda subordinates a passage to a function — the sentence was always doing a job, and the interpreter identifies the job. One asks what the text most deeply means; the other asks what the text is for.
The insight that neither generates alone is about accountability. A hermeneutic organised around depth has no natural stopping point, because there is always a deeper sense available and no external check on the descent. A hermeneutic organised around function does have one: the function is fixed by the injunction the passage serves, and if you cannot name that injunction you cannot claim the licence. That is why arthavāda's extension to free-floating Purāṇic narrative is not a small technical slippage — it removes the very anchor that made the device auditable, and it does so at exactly the moment the device is being used on the most contested material. A reader who wants to use arthavāda seriously has to supply the anchor the essay left out.
To Selective Honesty as Control Mechanism. Arthavāda formalises something that operates elsewhere as manipulation: the licensed distinction between what a statement says and what it is doing. In the behavioural-mechanics framing, controlling which register an audience reads you in is a power move — the speaker retains deniability about the surface claim while collecting its effect. Arthavāda is the same distinction, run in public, with the criteria published.
What the pairing produces is a test worth carrying: the difference between exegesis and manipulation is not whether a speaker separates saying from doing, but whether the criterion for the separation is stated in advance and applies symmetrically to passages the speaker likes. Dinanath's essay passes that test on the Vāyu episode — the criterion is named and prior — and the compendium as a whole fails it, since no essay asks whether guṇavāda might apply to the Rudrāṃśa claims it wants read plainly. The test is portable to any tradition, any doctrine, and any argument that leans on the distinction.
Sharpest implication. Arthavāda's real contribution is not permission to read figuratively; every tradition has that. It is the demand that the licence be published — three named categories, an external-contradiction requirement, and a subordination relation that can be checked. A tradition that reads figuratively without a published licence is doing the same thing with none of the accountability, and the published licence is what distinguishes interpretation from convenience.
Generative questions.