What This Concept Is About
"能知" is one of those Buddhist terms that, the moment you try to grab it, slips through your fingers—and that slipperiness is precisely the point.
In the most basic sense, 能知 (literally "that which is able to know") designates the cognitive subject-side of any act of knowing. Whenever there is knowing happening, there are two sides: the known (所知, jñeya) and the knower (能知, jñātṛ). When you see a blue sky, the sky is 所知, and the seeing-as-an-activity-coming-from-somewhere is 能知. When you recall your mother's face, the face-image is 所知, and the recalling is 能知. The term covers anything that performs cognitive function: 識 (vijñāna, consciousness), 心 (citta, mind), 慧 (prajñā, discriminating wisdom), 念 (manas, thought-attention)—all are ways of naming the "side that knows."
But here's the crucial twist, and the reason this concept is philosophically loaded rather than merely descriptive: 能知 cannot stand on its own. It is not a thing, an entity, a soul, a self, or a substrate that "has" knowing. It is defined entirely by its relation to 所知. Take away what is known, and there is nothing you could point to as "the knower." Change what is known, and the knower changes with it. This is the meaning of the central thesis from Tsongkhapa's commentary on Candrakīrti's *Madhyamakāvatāra* (《入中論善顯密意疏》):
「能知所知，俱無自性。」— "The knower and the known both lack self-nature."
And the even sharper version:
「實無少法能知少法。」— "Not even one phenomenon knows another phenomenon."
This is not a flat denial of cognition. Within the conventions of ordinary language (名言諦), knowing certainly happens: when consciousness engages with an object-form, when mind arises holding the appearance of a blue thing, when wisdom displays the characteristic of an ultimate meaning—in all these cases we conventionally say "knowing occurred." But none of these events can be located as a fixed, intrinsically existing "knower" standing behind the knowing. The knower is a designated function (假名施設), a label pinned onto a relational event, not a substance uncovered beneath the event.
The commentary lays out a specific logical path for establishing this. You cannot, on the one hand, point to some object-as-known-without-a-knower—that wouldn't make sense. You cannot, on the other hand, point to some knower-as-knowing-without-something-known—that wouldn't make sense either. The two are mutually conditioning (相待). So if you can show that the known has no self-nature, the knower falls with it; this is why the Madhyamaka tradition typically negates 所知 first, and the negation of 能知 follows as a consequence. The path is not "smash the knower" but "see through the relationality, and the knower's supposed solidity dissolves."
Āryadeva's *Śataśāstra* (《百論》) approaches the same territory from a very different angle—through debate with non-Buddhist schools (especially the Ājīvika/Sāṃkhya-type opponents who posit a permanent "soul" or *puruṣa*)—and arrives at a complementary conclusion. The opponent says: behind the changing stream of cognition there is a permanent soul (神) that does the knowing; when consciousness, mind, and object come together, the soul "uses" the cognition to know objects. Āryadeva's reply is devastatingly simple:
「但知能知，非神知。」— "Only knowing itself is the knower; it is not the soul that knows." 「知即能知。」— "Knowing itself *is* the knower."
If a knowing were not capable of knowing, we wouldn't call it knowing. So cognition is not a tool wielded by a soul; cognition itself is what knows. Positing a separate soul-knower adds nothing explanatory—it's like saying "the lamp sees" (the lamp only aids seeing; it doesn't see). The opponent tries analogies: "a horse-body and a black blanket together form 'the caparisoned horse'—so the soul and cognition together form the knower." The reply: a man yoked with an ox does not thereby become an ox, nor does the ox become a man; the compound doesn't transfer properties. The characteristic of knowing lies in knowing itself, not in the soul joined to it.
There's a second line of argument in 《百論》worth preserving in its full logic: the partial-knower refutation (分知不名知). If the soul is all-pervasive and cognition only arises when it meets a small portion of the soul, then what knows is only "a part" of the soul. But if you call partial cognition "knowing," then by the same logic you could call the other (much larger) part that doesn't know "unknowing"—and then the soul would be characterized as unknowing. You cannot use the same whole to claim both "knows" and "doesn't know" depending on which part you sample. The hand that writes doesn't make the whole body a writer; the garment that half-burns doesn't make the whole garment burned. The conclusion: "莫著語言"—"don't get fixated on linguistic labels."
Now, several refinements are important before we leave the technical core:
First, the commentary distinguishes 能知 from the Yogācāra-style "self-cognition" (自證) that some schools use. In some interpretations of consciousness-only, each moment of consciousness allegedly knows itself directly, with no subject-object split—pure internal self-luminosity. Tsongkhapa's commentary explicitly distinguishes the "mind that feels" (能受之心) from this alleged self-cognition: the feeling mind still has subject-object structure (it feels *something*, namely pleasure or pain), so it is not the undifferentiated self-knowing consciousness some posit. This means 能知, in this tradition, is not an endorsement of pure inner self-awareness; it is precisely a relationally-defined cognitive function that still has its "other" side.
Second, in the context of buddha-wisdom (佛智), the commentary gives us the most radical statement. The "wisdom of suchness" (如所有智) of a fully awakened buddha does not display the dual appearance of knower and known. Before awakening, a single moment of wisdom cannot grasp both a particular phenomenon and its ultimate nature simultaneously—they must be grasped separately. After awakening, the buddha abides continually in the fundamental concentration that directly realizes the ultimate, where the two-appearance structure of "something known" and "someone knowing it" is no longer displayed. So 能知, in its ultimate resolution, is not affirmed as an enduring subject; it is the conventional designation that *itself* dissolves into direct realization.
This is the full picture: 能知 is the name we give to the subject-side of cognition. It is real as a conventional designation but hollow as an intrinsic nature. It is relationally defined, not substantially grounded. And in the highest realization, even this relational designation is transcended—not by destroying knowing, but by knowing in a way that no longer fabricates a separate knower.
Walking Through It in Daily Life
This is where the philosophy gets feet. Let's take the most ordinary moment you can think of—you, on a train, scrolling through your phone. Walk it step by step, mapping carefully to the classical definition.
Step 1: The moment a notification appears. Your visual consciousness (眼識) takes the shape of the icon; cognitive consciousness (意識) recognizes it as a message from a particular person. Here, the message is 所知, the perceiving-and-recognizing activity is 能知. Already we see: there is no fixed "you" who is the knower, independent of the message. The "knower" right now is *the act of this recognition*—shaped by the message, by your history with this person, by the fact that you're tired, by the train noise. Change any of these and a different "knower" emerges.
This already troubles the common intuition that "I" am the knower and the world is the known. The classical definition tightens the screws: not only is there no *separate* knower, there is no knower *at all* as a self-standing thing. The "knower" of this moment is a relational event, like the "left side of this coin"—meaningful only in contrast to its other side.
Step 2: The mind wanders to dinner plans. Suddenly you're not seeing the phone screen anymore; you're imagining a restaurant. What's the 能知 here? It's the *imagining*-mind, complete with the restaurant's image. Notice how the "knower" has shifted: the previous moment's knowing-of-the-message is gone, replaced by knowing-of-the-restaurant. If you were looking for a fixed "I" that persists through these shifts, you wouldn't find one—only a stream of knowing-events, each with its own known-side. This is the lived experience of "能知所知相待而立"—the knower and known *arise dependently*. Neither is prior.
Step 3: A friend asks, "Are you even listening?" You realize you've been reading a menu in your head while they were talking. Here the famous "分知不名知" argument from 《百論》becomes very practical. You were "partially knowing"—knowing your thoughts but not knowing your friend's words. If the conventional "I" is taken as a unified knower, then in this moment it is both a knower (of the restaurant) and a non-knower (of the friend's sentence). The same "self" cannot coherently bear both characterizations at once. The "knower" that knows the restaurant and the "knower" that knows the friend's words are not the *same* knower—they are partial, momentary, function-defined. The whole-person-of-you that you ordinarily take as the unified subject of all this knowing is a retrospective label, not a discovered fact.
Step 4: A quiet moment—no objects at all. You close your eyes on the train, briefly without thought. Is there still a knower? In some sense yes: there's a bare "knowing of silence" or "knowing of drowsiness." In another sense no: when there's nothing being known, the relational term "knower" loses its referent. This is why the commentary says 「若慧不現境相，便不於境轉，不能說知彼境」—if wisdom doesn't display the characteristic of an object, it doesn't engage with an object, and you can't say it knows that object. Even the seeming "knower of nothing" requires something to be the known—otherwise the term has no application. This is the practical meaning of "實無少法能知少法"—no phenomenon knows another phenomenon as a self-existing entity knowing a self-existing entity.
Step 5: A moment of empathy with the friend. Now you *do* listen, and you feel what they're saying. The classical distinction between 能受之心 and mere cognition becomes relevant here. The mind that feels (受) is still relational—it feels *something*, it knows *something* about the other's situation. It is *not* a pure undifferentiated self-luminous awareness. To mistake the feeling mind for an inner "self that knows itself directly"—that is the trap of 自證. The commentary is warning us: even when the known seems internal (an emotion, a feeling), the knowing-of-it still has a structural two-sidedness. There's no escaping relationality into a subject-pure ground.
Step 6: A sudden question—"who is doing all this knowing?" The moment of inquiry itself. Here you can watch the 能知 trying to turn back on itself, looking for itself. The classical teaching is that this reflexive move doesn't find a self-standing knower—it finds only more knowing-events, each again relationally defined. The search for the "knower behind the knower" is exactly the move that, when examined, yields "能知所知，俱無自性."
The most common mapping error to avoid in this walkthrough: people often collapse 能知 into "the mind" or "consciousness" as if these were independent substances. The classical definition forbids this. "Mind," "consciousness," "wisdom" are all *terms for the knower-side of a knowing event*. They are not separate things that do the knowing; they are ways of naming the subject-function within a relational whole. The error is like saying "the left side of the coin is what makes the coin a coin"—misplacing the ontological weight onto one relational aspect.
Why Contemporary People Need This
Modern life is in many ways a project of cultivating, perfecting, and protecting a particular kind of "knower." Self-help culture, productivity literature, identity politics, even certain strands of contemplative mindfulness treat the knower as the thing to be optimized. "Be the observer of your thoughts." "Know thyself." "Cultivate your awareness." These are all good-faith projects that take 能知 as a real, cultivable, identity-bearing entity.
Buddhist analysis of 能知 does not dismiss these projects, but it does point to a structural blind spot in them. The blind spot is this: the knower you are trying to optimize is itself a relational, function-defined process—not a fixed center. So every act of optimizing the knower reinforces a subtle misdescription of what is actually happening. You end up with an ever-sleeker, ever-more-attentive "knower" that is still hollow at its core—not because nothing is happening, but because what is happening was never the kind of thing that has a core.
Why this matters for contemporary distress:
The deep anxiety of our age is, arguably, the anxiety of the knower. We are exhausted by endless self-monitoring ("how am I feeling right now?"), by constant orientation to how we are perceived ("what will they think of this post?"), by the impossible demand to be the unified, continuous, authorial subject of a coherent life-narrative. This exhaustion is not a failure of cultivation; it is the predictable consequence of trying to be a knower who cannot, as a matter of philosophy, be the kind of thing we are trying to be.
The Madhyamaka analysis offers something specific here: it doesn't say "stop knowing" or "kill the knower." It says: when you know, recognize that knowing is a relational event with no owner standing behind it. The knowing itself can be vivid, accurate, kind, skillful—none of these require that there be a substantial knower producing them. The 能受之心 that empathizes with your friend doesn't need to be grounded in a fixed self to be genuinely empathic. The 慧 that discerns truth doesn't need to be owned by a permanent knower to discern truly.
In terms of contemporary analogies (used as analogies, not as proofs): the modern neuroscience picture of "predictive processing" describes the brain as constantly generating predictions (a kind of knowing) about sensory input, with these predictions being continuously updated by prediction-error signals. There is no fixed "predictor" in this picture; there is a process of prediction-generation. The analogy to the Buddhist analysis is suggestive—both systems refuse to locate a central knower behind the knowing—but neuroscience has not and likely cannot deliver the same philosophical conclusions, because its questions are different from the Buddhist ones about ultimacy and conventional designation. The analogy should be held lightly.
Similarly, the way algorithms "know" us—curating feeds based on what we click—presents a disturbing mirror: a knowing that is entirely relational, with no knower inside the algorithm, and yet highly effective at shaping what shows up to be known. We are uncomfortable with this because we want our knowings to come from somewhere; the Buddhist analysis suggests they always were already of this character, even when a "self" seemed to be doing them.
The practical upshot is not nihilism but relief. When you no longer need to maintain the fiction of a fixed knower, a great deal of labor stops. The constant work of being a coherent self, of tracking your own states, of defending the knower's position—all of this turns out to have been optional, a kind of metaphysical overwork. Knowing can continue, vividly and accurately; it just doesn't need to be *yours* in the substantial sense.
Common Misreadings and Clarifications
Misreading 1: "能知 is being denied, so Buddhism is nihilism about the self." This is the most common error. 能知 is not flatly denied; it is denied as having self-nature (自性) while being conventionally affirmed as a functional designation. The commentary is explicit: in ordinary language (名言諦), knowing certainly occurs—consciousness engages objects, mind holds the form of the blue, wisdom can know ultimate meaning. These are not illusions to be discarded; they are designations to be understood correctly. The negation targets the *intrinsic nature* of the knower, not the functioning of knowing.
Misreading 2: "能知 is the true Self (ātman), and Buddhism is secretly teaching a Self." Some interpreters, especially those influenced by certain Vedāntic or Yogācāra readings, want to identify 能知 with a deeper, truer self that survives analysis. The Madhyamaka commentary explicitly blocks this move. The knower is relationally defined; it cannot be the foundation because it has no foundation. Any "self" you try to identify with 能知 will, on inspection, be either (a) another knowing-event (and thus not foundational) or (b) a verbal label pasted onto a stream of events (and thus not a discovered entity). 《百論》's entire argument is aimed precisely at this move—when the opponent says "the soul is what knows," the reply is "knowing itself knows; positing a soul-knower is explanatorily empty."
Misreading 3: "能知 is the same as consciousness-only's self-cognition (自證)." No. The commentary carefully distinguishes 能受之心 (the feeling mind, which still has subject-object structure) from 自證 (an alleged pure self-knowing without the two-sided appearance). Some Yogācāra and certain Tibetan positions posit a self-cognizing consciousness that knows itself directly, without being divided into knower-and-known. Tsongkhapa's commentary explicitly says the knower under discussion here is not this—even internal feeling and knowing still have their known-side. The point is sharp: if you mistake the relational knower for a self-cognizing ground, you've imported the very substantialism the analysis is meant to undo.
Misreading 4: "If there's no fixed knower, then everything is random and no one is responsible." This confuses the lack of a *self-natured* knower with the absence of a *functional continuum*. In conventional terms, the knowing-stream continues, conditions follow one another, intentions produce actions, actions have consequences. The analysis of 能知 does not erase these conventional facts; it analyzes their mode of existence. You can still be responsible within the conventional framework, just as the *designation* "responsible agent" can be meaningful even when no agent-substance is found beneath it. The commentary is at pains to preserve conventional cognition intact—it is the *ultimate* status of the knower that is being investigated, not its *ordinary* functioning.
Misreading 5: "能知 is a metaphysical entity described by Buddhism." This is the most subtle error and the one most worth watching. Even readers sympathetic to the Buddhist analysis slip into treating 能知 as a special kind of thing—a "Buddhist knower," a more refined object of metaphysical inquiry. But 能知 is not a thing described; it is a *term whose referent is exhausted by its relational role*. The analysis is showing you how to use the term correctly, not introducing you to a new entity. Once you grasp the relationality, the term has done its work; you don't then have a more sophisticated metaphysical object called "能知" to add to your ontological inventory. This is one place where the Buddhist analysis converges with certain strands of contemporary anti-foundationalist philosophy (Sellars, Brandom, even some Wittgensteinian moves)—but again, the convergence is analogical, and the Buddhist tradition has its own internal reasons for the position.
Misreading 6: "If 能知 has no self-nature, awakening means losing all capacity to know." This is contradicted directly by the commentary's treatment of buddhahood. The fully awakened wisdom (如所有智) doesn't *lose* knowing; it knows without the dual appearance of knower-and-known. This is a different *mode* of knowing, not an absence. Analogy (again, held lightly): when you become deeply absorbed in playing music, there isn't a separate "you watching yourself play"; there's just the playing. The musician-knower hasn't disappeared; the knower-known split has temporarily become irrelevant. The Buddhist claim is that this non-dual knowing is the natural mode of cognition when the fabricated knower-side has been seen through—not as suppression, but as the disclosure of how knowing always already was.
The deepest thing to take away may be this: 能知 is not a puzzle to be solved but a habit to be noticed. Every time you say "I know" or "I see" or "I feel," there is a tiny, almost invisible move being made—the move of positing a knower standing behind the knowing. The Buddhist analysis doesn't ask you to stop knowing. It asks you to notice, gently and repeatedly, that the knower you posit is a relational designation, not a discovered entity. In noticing, knowing doesn't collapse; it loosens. And in the loosening, something like clarity—unowned, ungrounded, vivid—can take its place.