Take a moment and ask: who knows that I'm reading this right now?
The answer feels obvious. Of course there is a knower — me, the one behind my eyes. The whole infrastructure of self-help, much of modern psychology, and most everyday speech is built on this assumption. The Buddhist tradition, especially in its Madhyamaka (中觀, "Middle Way") wing, picks this assumption apart with extraordinary care. The technical term for what gets picked apart is 知 (jñāna) — "knowing" or "the cognitive function." What follows is a careful, contemporary reading of this term, drawing primarily on two classical sources: 《中論》 (Mūlamadhyamakakārikā, *Treatise on the Middle Way*) and 《百論》 (Śataśāstra, *A Hundred Treatises*), both foundational Madhyamaka works that use 知 as a centerpiece of their analysis.
What This Concept Is Saying
The plain meaning
In its most direct sense, 知 is the cognitive function of mind — the activity by which a mind takes up an object (所緣, "the supported object") and registers it. When you hear a sound, recognize a face, or notice a thought passing by, that "registering" is 知 at work.
But here's the catch. The moment we say 知, the question "who knows?" is already pulling us toward a second, more troubling idea — 知者, the "knower." And once we have a knower, we are not far from a self (我), an ātman, a soul. The two Madhyamaka treatises in our tradition don't try to ground 知 in some new, better metaphysical foundation. They take the opposite route. They examine the assumption that 知 (and the knower it implies) is a stable, self-defining entity, and they show, layer by layer, that this assumption cannot hold up under inspection. The teaching isn't "there is no knowing" — knowing clearly happens. The teaching is "knowing cannot be owned by an unchanging knower."
Four layers of refutation in 《中論》
The 《中論》 does not attack 知 from a single angle. It opens four different lines of questioning, each aimed at a different way we might try to fix 知 in place. All four need to be kept distinct — they are not variations on one argument but four independent pressure points.
Layer One — Being, non-being, and the knower of both.
Earlier in the text, being (有, "existence") and non-being (無, "non-existence") have already been shown to be untenable as ultimate categories. The first move here is deceptively simple. If being and non-being are both refuted, what about the one who knows them? "若有知有無者，彼應在有中或在無中" — if there is a "knower of being and non-being," where exactly is this knower located? In being? In non-being? In neither? In both?
Each option collapses. A knower in being would itself be subsumed into the refutation of being. A knower in non-being would have no referent to know. A knower straddling both inherits the incoherence of both. The conclusion is not a nihilistic "there is no knower" but a precise diagnostic: wherever you try to lodge the knower, the location itself turns out to be untenable. This first layer establishes that 知 cannot be anchored in the most basic ontological categories.
Layer Two — The senses and their objects.
The senses themselves — eye, ear, nose, tongue, body, mind (六根) — are clearly pieces of flesh and nerve with no deliberation in them. "諸根無思惟，不應有知" — the senses have no thought, so they should not have knowing. But we obviously do see colors and hear sounds. So how does cognition arise?
An opponent — likely a school positing a unitary self — suggests: maybe beyond the senses there is a separate "knower of objects" (能知塵者). The 《中論》 then turns this proposal over twice. Is this knower located separately in each sense (so each eye has its own little knower, each ear another)? Or is the same knower spread across all senses?
The first option breaks down because we don't actually have six little knowers — we have one continuous visual field, one continuous stream of hearing, and these do not share their contents in the way the proposal implies (your eyes and ears do not exchange data the way phones do). The second option breaks down because a single knower spread across multiple organs would have to migrate between modalities, and we have no evidence of such an inner agent. The senses, the text suggests, are themselves implicated in cognition without being its owner. The point of this layer is not to deny that we see and hear; it is to deny that we can find a separate, definable "one who sees" anywhere in the apparatus.
Layer Three — The four-condition model.
Many schools in classical India proposed a model of cognition that looks strikingly modern: an "I" (我), a "mind" (意), a sense organ (根), and an object (塵) — four conditions — come together, and cognition (知) is "born." The 《中論》 examines when exactly this birth is supposed to happen.
If knowing arises *after* the object has been seen — first the bottle, then the knowledge of the bottle — then by the time you know it, you have already seen it; what additional work is the knowing doing? If knowing arises *before* the object is seen, the four conditions have not yet come together, so knowing cannot have been produced by their union. If knowing arises *at the same moment* as the seeing, the order the model itself required is lost: bottle first, seeing second, knowing third.
The conclusion is striking: "知無故，見、可見、見者亦無" — because knowing has no such origin, neither does seeing, nor the visible, nor the seer. The whole structure unravels. And rather than leaving us with nothing, the text offers a different metaphor for what is happening: "諸法如幻如夢，無有定相" — all phenomena are like an illusion, like a dream, without fixed characteristics. The knowing is happening, but it is not a thing; it is an event without an owner.
Layer Four — The self as knower.
The fourth layer tackles the most stubborn version of the knower idea: the proposal that there is a "me" (我) who knows, while my body does not. "I have knowing, the body has no knowing — therefore, the knower must be me." This is the reasoning of someone who feels consciousness is a non-physical property belonging to a soul.
The 《中論》 will not accept the inference. The proposed "me" is formless, intangible. A formless, intangible thing cannot act directly (自作), nor can it make something else act (使他作). It cannot be in any one place doing the knowing. And then come the very concrete counter-examples: my eyes see colors but cannot hear sounds — if a unitary "me" inhabited both senses, why does my eye's seeing not show up in my ear's hearing? I can write with my right hand, but the same skill does not transfer to my left unless I practice it. Watching someone else eat a sour fruit can make my mouth water — but this is not my mouth actually tasting; it is memory or imagining at work. In each case, the supposed "all-knowing, all-acting" self turns out to be unnecessary; cognition can be explained by the organs, the conditions, and the patterns — without any homunculus sitting at the top.
The text concludes with a line that deserves to be remembered: "離身心，更無別作者" — apart from body and mind, there is no separate doer. By implication, apart from body and mind, there is no separate knower either.
《百論》's distinctive contribution: "knowing is itself the knower"
The 《百論》 takes a different angle. Instead of laying out four angles of refutation, it stages a debate between an inner position (內方, the Buddhist voice) and an outer position (外方, various non-Buddhist schools — Vaiśeṣika, Sāṃkhya, and others). The central opponent is the idea of 神 — a permanent self or soul, the rough equivalent of ātman — that supposedly uses mind, senses, and objects to know things. The Buddhist voice offers a simple but radical claim: 知即能知 — knowing is itself that which knows. There is not a 神 behind knowing, waiting to use it as a tool. Knowing does not need an owner any more than (or rather, less than) a lamp needs a separate "lighter" to give light. This lamp analogy is famous in the Indian tradition; the 《百論》 says, pointedly, that the lamp analogy "does not work" when the opponent tries to use it to support 神 — it actually undermines 神.
This argument is unpacked through several distinct moves.
Against "神 and 知 are different but combine." The Vaiśeṣika-style opponent argues that 神 and 知 are different but combine, so 神 becomes the knower. The inner voice replies: the *characteristics* of knowing — discriminating, recognizing, taking up objects — those are properties of knowing itself, not properties of 神. So even if 神 and 知 "combined," 神 does not thereby become the knower. "知相住於知中，非神中" — the mark of knowing resides in knowing, not in 神. The text uses the same analogy the opponent had been using, of a person merging with an ox ("人與牛合") — and turns it back: if a man and an ox combine, the resulting entity has the ox's characteristics because of the ox, not because of the man. So if 神 and 知 combine, the resulting entity's knowing comes from 知, not from 神.
The part/whole problem. The Sāṃkhya-style opponent argues that 神 knows in parts but is still called "the knower" — a single self whose knowing is distributed across many organs. The 《百論》 turns the part/whole logic against it: "分知不名知" — partial knowing is not knowing. If 神 only knows in part (in each sense organ), then 神 in itself is not a knower. The text uses the very analogy the opponent would have used — a piece of cloth burning. Burning part of a robe does not burn the robe; many parts unburned does not make the robe unburned. So knowing in part does not constitute knowing by 神. The same logic is applied to 念 ("memory," "recollection"): if memory is divided across the senses (left eye sees, right eye does not), then memory belongs to the senses, not to a 神 above them. The opponent's "念由神中生" — recollection arises from 神 — does not hold up.
The conditions for knowing, again. The 《百論》 also subjects the conditions for knowing to the same pressure the 《中論》 does. If eye, object, and mind together produce knowing — when does this happen? Same three-way bind as before: after, before, or at the same moment. None holds ("生無生共不一時生"). And a particularly striking empirical observation: "近瓶遠月一時見故，知眼不去" — if you can see a near bottle and a far moon at the same moment, the eye is not traveling to the object. So the "eye goes to meet the object" version of knowing fails, and with it the union model. (When the senses do not combine, there is no union; when they are said to combine but cannot, the model collapses.)
Where this sits in the path
It is worth pausing on something easy to miss. The two treatises do not replace 知 with something else. They do not say "actually, what you call knowing is really X (some deeper essence)." They do not propose a "true knower" hidden behind the apparent one. They do not even deny that knowing happens. What they do is refuse to let us anchor knowing to a fixed subject — be it being, the senses, the four conditions, or the ātman. Knowing is real; the knower as an entity is not.
This puts 知 in a very particular place in the Buddhist map. It is *not* 識 (vijñāna, "consciousness") — the Yogācāra and Abhidharma traditions have a sophisticated vocabulary about consciousness, its eight kinds, its arising and ceasing. It is *not* 心 (citta, "mind") — the term that carries so much weight in Tathāgatagarbha ("Buddha-nature") thought. The choice of 知 is deliberately bare: it picks out the bare fact of cognitive occurrence and refuses to dress it up in metaphysical clothing. The two treatises use 覺 ("awareness") and 知 interchangeably, which is itself a clue — they are naming a function, not a substance.
In the larger map of the path, this analysis belongs to the dismantling of self-view (我見), specifically the variant that takes the form of "I am the one who knows." Without dismantling this, the deeper insights of emptiness and the realization of liberation remain incomplete. So 知 sits at a load-bearing point in the structure: not glamorous, not the heart of the path, but indispensable for anything that comes after. The famous "照見五蘊皆空" of the 《心經》 — "see the five aggregates as empty" — is the sweeping statement; the 知 analysis is one of the careful pieces of inspection that makes that sweeping statement intelligible.
Walking It Through in Contemporary Life
The classical arguments can sound abstract. But the moment you bring them into a familiar scene, they become very concrete — and a bit unsettling.
Scene one: the morning scroll.
You are on your phone. The screen shows you a video. For a moment, you stop and notice you are watching it. The reflexive move is to say "I am watching this." Who is watching? The classical Indian opponent in the 《百論》 would say: that is 神 — your true self — operating through the eye and the screen. The 《中論》 would say: wait, can you actually find that watcher?
Try it. Look for the watcher. Is it behind the eyes? But the eyes are just pieces of the visual apparatus. Is it in some inner screen where the picture appears? That inner screen is not anywhere you can locate. As soon as you try to find it, the looking itself becomes another object — and there is no homunculus behind it. You do not end up finding a self; you find an ongoing looking. 知 is happening. 知者 is nowhere to be found.
The mistake to watch out for: it is tempting to think "well, there is clearly *something* doing the watching." But "something doing the watching" is just a description of what is happening — it is not a separate entity. The watcher is the watching.
Scene two: the argument.
You are in a disagreement with someone you love. It goes badly. Later, replaying it, you say: "I knew what I meant. They didn't see it." The whole structure of the argument rests on a "me" who knew clearly, a "they" who failed to know, and a moment of knowing that belongs to me. The 《中論》's four-condition analysis is almost a textbook description of what just happened: you had a thought (your meaning), your partner's words came as an object (the object of hearing), your intention was to communicate (the mind), and somewhere in that, "knowing" arose. But did the knowing come after the words? Before? At the same time? You cannot actually pin it down. And when you try to ask "but I really did know, didn't I?" — the 知 was real, the content was yours, but the "I" doing the knowing cannot be located as a separate thing.
The mistake to watch out for: confusing "I had a real experience" with "there must be an 'I' who had the experience." The first is true. The second adds an entity that does not help explain anything.
Scene three: the meditation cushion.
You sit down. You try to "observe" your breath. The instructions say "watch the breath." Now: who is watching? This is precisely the question the treatises are built around. In some meditation traditions, the answer is: there is a "witness" (the inner observer) — a subtle self that watches everything else. This is a position the 《中論》 and 《百論》 are explicitly dismantling. The "witness" does not show up as a discoverable object. When you look for it, you find only the looking. The witness is a useful working metaphor in some practices, but if it is taken as a real entity, the teaching is being undone.
The mistake to watch out for: turning the meditation instruction into a metaphysical claim. "Watch the breath" can be a powerful instruction for settling attention. It does not entail that there is a watcher.
Scene four: the algorithm.
Now look outward. Your streaming service "knows" what you will like next. How? There is a system: a model trained on data, weights adjusted through gradient descent, predictions made through pattern matching. The system produces knowing — predictions, recommendations, classifications — that are often more accurate than your own. But there is no "knower" in this system. There is no place where the knowing "lives." It is distributed across parameters, hardware, and training process. No single location says "I know what this user wants."
This is the kind of distributed, ownerless cognition the 《百論》's 知即能知 is gesturing at. Knowing without a central knower is no longer just a philosophical puzzle; it is an engineering reality.
*Important caveat*: this is an analogy, not an identity. The way algorithms produce outputs is very different from how human cognition unfolds. Algorithms do not have first-person experience; they do not suffer or rejoice at their predictions. The analogy is meant to show how *knowing can occur without a knower*, not to claim the two are the same thing.
Scene five: the grief.
Someone you love dies. You say, "I know they are gone." That knowing is the most real thing in the world to you. The 《中論》 is not denying it. What it is diagnosing is the reflexive add-on: "and the one who knows this is *me*, a continuous me, a me that has always been the knower and will go on being the knower." In the rawness of grief, that add-on can actually be a comfort — there is a "me" who grieves, and that me will continue. The teaching is not asking us to drop that comfort in a moment of pain. It is asking us, much later, on quieter days, to look more carefully at whether the add-on is accurate. It is possible to honor the knowing without committing to a permanent knower.
The mistake to watch out for: treating the Madhyamaka analysis as an instruction to suppress "I know" language. It is not. The ordinary language of knowing is fine. What is being examined is the metaphysical claim that this language points to a stable, independent entity.
Why Contemporary People Need This
A few words on why this old analysis is not a museum piece.
The self-discovery industry is built on a homunculus.
Much of contemporary self-help, much of "find your true self" culture, much of personality psychology up until quite recently, has been built on the assumption that inside each of us there is a stable knower — a "real me" — and the project of life is to align the surface self with this inner one. The Madhyamaka analysis in 知 directly contradicts this. There is no inner knower to find. There is a flowing process of knowing, and aligning yourself with it is a very different project from "finding your true self." Many people have spent years, and considerable sums, chasing a self that the tradition says is not there to be found.
Mindfulness culture has begun to reify the watcher.
As mindfulness has moved from Buddhist contexts into corporate offices and clinical settings, it has sometimes carried with it, almost without noticing, the very reification the Madhyamaka treatises warned against. "Observe your thoughts" becomes, in some popularizations, "find the observer." "Be aware" becomes, in some renderings, "tune into your awareness-self." The 《百論》's 知即能知 is a useful corrective: you don't need to find an observer to observe. Observing happens. The activity is enough. This is not a small thing for people who sit in meditation hoping to find "their true self" and feel they have failed when they cannot locate it.
Neuroscience and AI keep finding no center.
Modern neuroscience has searched for the "I" in the brain and not found one. The sense of self is constructed by distributed networks; there is no CEO neuron, no single self-region. Modern AI has shown, in a different domain, that sophisticated knowing-like behavior can emerge without any central knower at all — distributed systems that produce knowing without a knower. Neither of these scientific findings "proves" the Madhyamaka analysis. But they make it easier to hear. A claim that would have sounded wild in the fifth century — that knowing has no owner — sounds far more plausible when contemporary science itself keeps failing to find an owner.
*Important caveat*: these are partial resonances, not identities. The Buddhist analysis is operating at a different level — it is about lived experience and the possibility of liberation, not about neural correlates or computational architectures. Using science to "prove" Buddhism is a well-meaning but conceptually slippery move. The two traditions can illuminate each other; they should not be conflated.
The relief of not having to maintain a self.
For some people — particularly those who struggle with the relentless project of being "themselves" — the Madhyamaka analysis offers an unexpected kind of relief. If the knower is not a fixed thing, then the exhausting work of being a coherent, continuous self does not have to be performed as urgently. You can have experiences, including painful ones, without the additional burden of constructing a "me" who owns them. This is not escapism; it is a different understanding of what a person is.
Common Misreadings and Clarifications
Given how much depends on getting this right, it is worth being explicit about the ways 知 is regularly misread.
Misreading 1: "知 means consciousness, and the teaching is about pure consciousness."
Not quite. The Madhyamaka analysis is not saying 知 is some purer consciousness hidden behind ordinary consciousness. There is no "pure consciousness" being proposed as the alternative. 知 is just the bare cognitive occurrence — what happens when something is registered. The two treatises refuse to let us elevate that occurrence into a substance, even a spiritual one.
Misreading 2: "If there's no knower, then nothing is really known. This is nihilism."
This is the most common objection and the one the texts are most careful to avoid. The refutation of the knower is not the denial of knowing. Knowing happens — colors appear, sounds register, ideas occur. What is refused is the addition of a stable entity to which the knowing belongs. The texts are dismantling a metaphysical extra, not the experience itself. The illusion metaphor in 《中論》 ("諸法如幻如夢，無有定相") is meant to be reassuring: the dream is real as experience; it just has no owner.
Misreading 3: "知 is the true self in disguise."
Some popular renderings of Buddhist thought slide, perhaps inadvertently, into a position that looks remarkably like Hindu ātman: there is a deep, true "knower" inside, and the project of practice is to realize it. The 《中論》 and 《百論》 are explicitly working against this. The Sāṃkhya-style opponent in 《百論》 is exactly this view — a 神 who knows through the apparatus — and the entire text is structured to dismantle it. 知即能知 is not "knowing has a deeper essence"; it is "knowing is itself the knowing."
Misreading 4: "This denies that 'I' exist or that we have agency."
The two treatises are dismantling a specific kind of self — a permanent, owner-of-experience self — not the functional, conventional self that plans breakfast, holds a job, and answers to a name. In ordinary language, "I know" is perfectly fine. Conventional selves operate, choose, and bear consequences within a Madhyamaka framework without contradiction. The teaching operates at a different level. It does not say "I don't exist"; it says "the kind of 'I' I'm usually assumed to be — the knower who owns experiences — isn't an accurate description of what's going on."
Misreading 5: "The watcher found in deep meditation is what 知 really means."
Many contemplatives report a sense, especially in deeper states, of a stable witness behind experience. This experience is real and has its own value. But the 《中論》's analysis is precisely aimed at the metaphysical reading of this experience. There is no permanent witness to find; there are moments of apparent witnessing, which themselves arise and pass according to conditions. To turn the witness-experience into a self is the move the texts refuse.
Misreading 6: "This is the same as Western process philosophy or the bundle-theory no-self of David Hume."
There are family resemblances, certainly. Hume's bundle theory — the self is just a bundle of perceptions without a single underlying substance — sounds similar in some ways. Buddhist process thought, especially in early Nikāya texts, has its own bundle-like accounts. But the Madhyamaka analysis in 知 is doing something more radical than bundle theory: it is not just saying the self is a bundle of perceptions; it is saying the perceptions themselves do not have the kind of fixed self-existence that bundle theory might still implicitly grant them. The "no knower" is part of a broader "no underlying essence in anything." This is a deeper cut than most Western no-self philosophies achieve, and the practical and liberative stakes are different.
Misreading 7: "Modern neuroscience proves Madhyamaka was right all along."
No. Neuroscience finds no central self in the brain; Madhyamaka finds no self in experience. These are different claims about different domains. The convergence is suggestive and useful, but it is not proof. Neuroscience could, in principle, discover something it currently calls a "self-area" tomorrow (this would be surprising, but the history of neuroscience has surprises). The Madhyamaka claim does not rest on any neuroscientific finding. It rests on the phenomenological and logical analysis found in texts like 《中論》 and 《百論》.
A Final Word
The question "who knows?" feels like a question with a definite answer. The tradition in these two treatises suggests the question itself is built on a subtle error. There is knowing. There is no fixed knower waiting behind the knowing to be discovered. Recognizing this — gently, gradually, not as a metaphysical thesis but as a lived insight — is one of the load-bearing insights of the Buddhist path.
It does not require you to stop using the word "I." It does not require you to deny your experience. It asks only that, in moments when you can, you look for the knower the way 《中論》 invites you to look — in being, in non-being, in the senses, in the four conditions, in the self — and notice that, despite all the looking, no separate knower turns up. What is left is the knowing itself: open, ownerless, ongoing.
That is not a nihilistic emptiness. It is the actual texture of experience, finally seen without the metaphysical addition.