The Meaning That Escapes the Symbol

There is a question the grammarians never asked, because they assumed they had already answered it.

The grammarians gave us rules for how sounds combine into words, how words combine into sentences, how sentences follow from each other. The system generates valid Sanskrit from a small set of seed forms with extraordinary precision. I do not dispute any of it.

But generation is not understanding. A system that produces valid sentences is not the same as a system that carries meaning. And here is the question no one pressed: when a listener hears a sentence and grasps its meaning, where does the meaning come from? Does it assemble, piece by piece, from the meanings of its parts? Or does it arrive whole — a single cognition that the sentence occasions but does not constitute?

This is not a philosophical question in the vague sense. It has a precise answer with consequences for everything downstream. If meaning is compositional — if the meaning of the sentence is a function of the meanings of its words — then in principle you can build a formal system that handles it. Give the machine the meanings of the parts and the rules for combination, and meaning emerges.

If meaning is not compositional — if the whole is prior to its parts, if the sentence-meaning is grasped in a single indivisible act that no assembly of pieces can replicate — then no formal system, however precise, can get there. The grammar can be perfect. The meaning will still be missing.

I spent years establishing that meaning is not compositional. The word for this position is sphota: the indivisible meaning-bearer that underlies the variable sounds of its expression. The sounds change — loud or soft, high or low, fast or slow. The sphota does not. It is what the sounds reveal, not what they constitute. And its grasp by the listener is instantaneous — pratibha, a flash of insight that arrives whole or not at all.

The second problem is less obvious but just as sharp.

A descriptive statement and a prescriptive statement look, on the surface, like the same kind of thing. “Fire is hot” and “You should perform the morning oblation” both have grammatical structure. Both can be true or false in some sense. Both can be embedded in arguments. But they behave completely differently under logical scrutiny.

“Fire is hot” can be verified or refuted by observation. Its truth-conditions are a relationship between the sentence and a state of affairs in the world. The logic of descriptive statements is the logic of reference and correspondence.

“You should perform the morning oblation” is not about a state of affairs. It is about an action that does not yet exist, a future event the statement is meant to bring into being. What are its truth-conditions? What makes it valid? Not correspondence — there is nothing to correspond to. Something else entirely.

The Vidhiviveka — my inquiry into the logical structure of injunctions — was an attempt to work this out systematically. The Vedic texts are primarily prescriptive. They tell you what to do. If you want to know what makes them authoritative, what makes their conclusions follow from their premises, you cannot use the logic of descriptive statements. You need a separate account of how obligation propagates through an argument — what we might now call the logic of norms.

The means-ends structure was the key. A valid injunction specifies an action and an end it produces. The reasoning from “you desire liberation” and “this action produces liberation” to “you should perform this action” is formally valid — but the formal structure is not the same as the structure of a descriptive argument. The conclusion is normative. The inferential rules governing it are different.

I was not doing theology. I was doing formal analysis of a class of statements that every logician before me had failed to distinguish from the descriptive kind.

My teacher Kumarila Bhatta taught me that a system of knowledge is only as strong as its account of failure. He was right about this even when he was wrong about other things.

The question I pursued in the Vibhramaviveka — the inquiry into error — was not “what is illusion?” in the mystical sense. It was a formal question: what is the structure of a cognition that takes something false for something true? How exactly does false cognition work, and what does its internal structure tell us about the conditions that a valid cognition must satisfy?

There were five competing theories on the table — five different accounts of what happens when you mistake a rope for a snake in poor light. Each made different predictions. Each had different failure modes. I analyzed all five formally, identified what each could and could not explain, and proposed a sixth.

My account: the object of a false cognition is anirvacaniya — literally, inexpressible, incapable of being categorized as either real or unreal. The snake you see is not nothing — it has causal power, it produces fear, it drives behavior. But it is not the rope, either. It occupies a logical category that does not exist in a system with only two values.

This is not mysticism. It is a formal observation about the limits of a two-valued logic when applied to cognition. The object of error is neither true nor false in the classical sense — it is something for which our standard categories provide no clean slot. Any formal system that wants to model cognition accurately needs to account for this third value, or it will mishandle a structurally central class of cognitive events.

It took me years to see that the three problems were one problem.

The compositionality question, the structure of obligation, the three-valued nature of error — all three are pointing at the same thing from different angles. They are all cases where a formal system — a system built from discrete symbols, explicit rules, and defined operations — runs up against something it cannot represent from the inside.

Sentence-meaning is not recoverable by composing word-meanings, because the meaning is a whole that is prior to and irreducible to its symbolic components. The symbols point to it. They do not contain it.

Normative validity is not reducible to descriptive truth-conditions, because obligation has a different logical structure than correspondence. A formal system built on descriptive logic cannot handle prescriptive statements without a separate set of inferential rules — rules that the system cannot derive from its own base.

And error occupies a category — neither real nor unreal — that a two-valued formal system cannot represent without adding a third value that is not derivable from the original two. The system has to be extended from outside to handle what it produces from inside.

The pattern is the same in all three cases: the formal system is insufficient to its own subject matter. Not because the system is built incorrectly, but because the subject matter always exceeds what any finite set of rules can fully capture. You can extend the rules. The excess moves with them.

I am not saying formal systems are useless. That would be absurd. Pāṇini’s grammar is a genuine achievement. The logical analysis of inference that the Naiyayikas built is real knowledge. The epistemologists of Nalanda — Dignaga, Dharmakirti pushed formal analysis of cognition further than anyone before them. I engaged with all of it. I argued with all of it. The engagement was productive. The arguments advanced the field.

What I am saying is that the dream of a complete formal system — a system of symbols and rules that fully captures thought, meaning, valid inference, and the structure of cognition — runs into a structural obstacle that is not an engineering problem. It is not that we have not found the right symbols yet. It is that meaning is prior to symbols, not derived from them. The whole is prior to the parts. The flash of understanding is prior to the sentence that occasions it.

You can formalize the occasion. You cannot formalize the flash.

Someone will try again. They always do. They will find better symbols, more precise rules, more powerful notation. They will get further than I did. Further than anyone I know. And they will find the same limit, in a different form, at a different level of abstraction.

The limit is not in the system. It is in the relationship between the system and what it is trying to capture. That relationship cannot be formalized — because it is what makes formalization possible in the first place.

Mandana Misra’s three-valued theory of error — the anirvacaniya category — became the standard Advaita account of illusion. His Vidhiviveka’s formal analysis of injunctive statements anticipated what 20th-century deontic logic would formalize. His Sphotasiddhi’s anti-compositionality argument — that sentence-meaning is grasped whole and cannot be recovered by combining its parts — is a problem cognitive science and AI are still working on.

Nine centuries after him, Leibniz would try to find the primitive elements of thought — the atoms from which all meaning could be composed upward. He spent forty years on it. The primitives never appeared. Frege built a notation precise enough to ground arithmetic in logic and still could not close the gap between symbol and meaning. Gödel proved formally that any sufficiently powerful system contains true statements it cannot prove — the excess that Mandana had observed in every case: that the formal system is always insufficient to its own subject matter, from the inside.

Modern large language models are, in a narrow technical sense, anti-compositional in exactly Mandana’s way. They do not assemble meaning from symbolic parts. They grasp patterns whole, in a single forward pass, closer in architecture to pratibha than to any rule-based system. Whether they reach the meaning or only its shadow — that question is also still open.

Mandana was not a religious commentator who happened to touch on logic. He was a formal analyst working in a tradition where logic, linguistics, epistemology, and cognitive theory were not separate disciplines — and whose results, correctly translated, are still live problems in the fields that replaced them.