这个概念在说什么
When contemporary readers see the word 假想, they usually translate it in their head as "imagination" or "fantasy," and the conversation is over before it begins. The 《瑜伽師地論》 (Yogācārabhūmi-śāstra, T1579) is using the term in something close to the opposite way. 假想 (Sanskrit *kalpanā*) there is not the airy drift of an idle mind; it is the technical name for the way names, perceptions, and speech are *laid over* a thing or an aggregate so that the thing can be talked about, identified, and reasoned with. The 論 places 假想 among the 心不相应法 (factors non-associated with mind) and defines its job in a single phrase: 「於諸事中為起言說」 — "it is that by which speech is initiated among dharmas." Read this carefully. 假想 is the bridge through which raw phenomena (諸事) get picked up, bundled, named, and made speakable.
The character 假 in 假想 is the most easily mishandled word in the whole discussion. It does *not* mean "false" in the English sense, nor "made-up from nothing," nor "delusion to be dismissed with contempt." 假 here is a technical marker that the thing referred to by a name *has no independent self-nature apart from its conditioning aggregate*. The Yogācāra author is not denying that the underlying dharmas exist; he is denying that the named object — once we have named it, drawn its outline, given it a noun — sits anywhere except as a stable imprint overlaid on a shifting cluster. When the cluster breaks up, gets recombined, is processed, or transforms, the old name is shed and a new one can arise in its place. This "shed-and-replace" behavior is exactly what marks 假.
The 論 builds the concept in five interlocking moves, and a serious reading has to keep all five distinct:
1. 假想 as the linguistic-perceptual installation of names. The 論 notes that 假想 cannot be sorted into the usual oppositions — 有色 / 無色 (form / formless), 有見 / 無見 (visible / invisible). It belongs instead under 安立道理 (the "principle of conventional establishment"). The reason is structural: 假想 is the operation by which a 事 (a thing-event) gets framed so 言說 can run on it. Without this installation, language would have nothing to bite on. So 假想 is not a side-effect of cognition; it is the apparatus that makes any proposition possible at all.
2. 想 (saṃjñā / perception) takes 言說相狀 as its object. This is the hinge that connects 假想 to the broader Buddhist analysis of experience. In the classical analysis, each of 識 (consciousness), 作意 (attention), 觸 (contact), 受 (feeling), 想 (perception), 思 (intention) has a distinctive object-characteristic it seizes. 想's distinctive object is 言說相狀 — the "linguistic shape" of a thing, the way a thing already wears the form of how it would be said. Because 想 is what grasps that shape, 想 becomes the doorway through which 言說 enters the mind, and 假想 is what gets built on the other side. To put it plainly: *every aggregate we call a noun was, at some point, a perception that had already been told what it would sound like*.
3. 假相有法 — the objects so designated. The 論 then makes a famous move. Over the five aggregates (色受想行識) we erect various "我" (self) theories. Over the assembled combination of 色香味觸 (form, odor, taste, touch) we erect the names 宅舍, 瓶, 衣, 車乘, 軍, 林, 樹 — house, vase, garment, vehicle, army, grove, tree. Whatever is talked about by these superimposed linguistic-perceptual installations is called 假相有法 — "phenomenally-posited existents," things that hold their shape only within a certain range and only while their underlying aggregate continues to cohere. When the aggregate is broken, recombined, processed, or transformed, the names are abandoned and new ones may be assigned. This is the diagnostic signature of 假有 (conventional / imputed being): it is real *within* its conditions, and it shifts as the conditions shift.
4. 假想物 vs 實物有 — what gets imputed vs what pervades. This is one of the cleanest contrasts in the treatise. The 論 asks: where do the various 想 of "food, vehicle, vessel, clothing, ornament" arise? Only in *specific* clusters of 色香味觸 — this dish, this car, this vase, this dress, this ring. Take those aggregates away or substitute others, and these 想 do not arise. By contrast, 想 of 色, 香, 味, 觸 themselves — the raw percepts of form, odor, taste, touch — pervade *all* of the above without distinction. So the 論 classifies the first group as 假有 (imputed being) and the second as 實物有 (real-object-being). The point is *not* that food is illusory and color is rock-solid. The point is that food can only ever be *this* assembly of percepts, while color, odor, taste, and touch show up across all such assemblies, indifferent to which assembly is in view. The "假" is not unreality; it is *dependence on a specific assembly*.
Within 假想物, the 論 further distinguishes two subtypes: - 依止一聚 (based on a single cluster): individual food, individual vehicle, individual garment, individual ornament. - 依止非一聚 (based on multiple / non-single clusters): paintings, sculptures, houses, palaces, armies, groves — things whose identity depends on *several* component aggregates coordinated together.
This second subtype is crucial for modern life, because most of what we care about — relationships, brands, organizations, careers, nations, reputations — is non-single-cluster.
5. 客法 and 客想 — names as the "guest" added onto things. The 論 drives the point home with a striking argument. If a thing were truly itself as the name says it is, and if the name were a faithful direct description of it, then *direct perception of that thing should not have to wait for having heard and intellectually distinguished its name*. But in fact, when we first encounter a dharma, the relevant perception does not arise properly until the name has been learned and the distinction has been drawn. Therefore, the 眼名 (the name "eye"), the 眼想 (the perception "eye"), and the 假施設 (the conventional imputation "eye") are all 客法 / 客想 — guest-dharmas, guest-perceptions, added onto their object rather than discovered within it. The same holds for ear, nose, tongue, body, and even for 見聞覺知 (seeing, hearing, sensing, knowing) themselves. The argument is austere: every linguistic-perceptual designation in our experience is, in this sense, a guest.
6. 假想 as a practice object. Having diagnosed the structure, the 論 turns the analysis into a修行 task. The practitioner is to 攝受能除遣所有假想的加行道 — undertake the preparatory path that is capable of removing all 假想. In every perception of dharmas, one is to generate 唯客想 — the recognition that "this is only a guest-perception." One is to 数数除遣一切戏论之想 — repeatedly dispel every 戲論之想 (discursive-proliferative perception). And one is to abide with 無分別、無相之心 (a non-discriminating, signless mind) that 唯取其義而轉 — moves by taking only the *meaning*, not the conceptual overlay. This is the precise moment when the otherwise dry analysis of names becomes a liberating operation. The 假想 analysis is not philosophy; it is the technical preparation for a 心 (mind) that does not rest on the superimpositions of speech.
Finally, the 論 uses a whole cluster of near-synonyms and they are *not* interchangeable: - 假想 — emphasizes the function of initiating speech about dharmas. - 假施設 — emphasizes the act of establishment by name and concept. - 假有 — emphasizes that the *named object* lacks independent self-nature. - 假相有法 — the technical term for such a named object within its range. - 假想物 — an object established by 假想, especially the assembled 宅, 軍, 林 type. - 假有想 — the *perception* itself as imputed, not as bare cognition. - 客法 / 客想 — the name and perception seen as the "guest" added onto the thing. - 戲論之想 — the proliferative perception that 修行 repeatedly dismantles.
To flatten all of these into "imagination" is to lose the entire operative vocabulary of Yogācāra. It also helps to mark here that the later 唯識 (Consciousness-Only) tradition crystallizes this analysis into its famous three natures: 遍計所執性 (imagined nature — what 假想 builds), 依他起性 (dependent nature — the aggregate of conditions on which it depends), 圓成實性 (perfected nature — what is realized when the imputation is seen through). The 《瑜伽師地論》's discussion of 假想 is, in important respects, the working ground out of which that later technical scheme is carved. The mature terms are not yet in place, but the conceptual cuts are.
One important disambiguation up front: 假想 is *not* 想蕴 (the aggregate of perception) and not the 心所 (mental factor) 想. 想蕴 is the entire mass of perception as one of the five aggregates a self-ish being clings to (most famously in 《心經》's line 「照見五蕴皆空」 — "clearly seeing that the five aggregates are all empty," a teaching that points precisely at the 假想 machinery woven into 想). 假想 is the narrower technical operation by which *names are laid onto aggregates so that language and identity-claims become possible*. It is one specific use of 想, not 想 itself.
生活里的走查
The classical example-set in the 論 was 宅舍, 瓶, 衣, 車乘, 軍, 林, 樹. To walk the same analysis in a contemporary apartment, commute, and phone screen, I'll keep the four distinctions sharp: - 實物有 — 色、香、味、觸 (form, odor, taste, touch) and their underlying moments, which pervade assemblies indifferently. - 假想物 (依止一聚) — a single-cluster imputation. - 假想物 (依止非一聚) — a multi-cluster imputation. - 客法 / 客想 — the name itself as added to its object.
And I'll flag at each step the place modern intuition most often slips.
1. Your morning coffee. You say "I'm having my morning coffee." 咖啡 as the named object is a textbook 假想物, 依止非一聚. It is not a single cluster. It is coordinated: the specific beans (their origin, roast), the grinder setting, the water chemistry, the device at home, the mug you reach for, the time of morning, the way the kitchen smells, the texture on your tongue, the small pause before the first sip. Remove or change any one of these aggregates — say, the machine breaks, the roaster changes their blend, you travel to a country where the same drink is made differently — and the name "my morning coffee" can no longer hold. The 假 has revealed itself. Meanwhile, what the 論 would call 實物有 — form, odor, taste, touch in their bare perceptive moments — continues to function in every other assembly of bitter aromatic liquid, including a pour-over at a friend's place that you would never call "my coffee." That is precisely the asymmetry the 論 points to. The *most common mis-mapping* here is to read 假想 as "fake coffee" — to think the 論 is claiming your coffee is unreal. It isn't. The 論 is claiming that "my coffee" is an overlaid name covering a moving cluster, and that the moments of 香味觸 themselves are doing their work in many other clusters too. A second mis-mapping, subtler and more harmful: reifying "my coffee ritual" until it quietly rules your morning and begins to extract suffering when the cluster shifts (illness, travel, loss of the mug, loss of the partner with whom you used to drink it). That reification is 戲論之想 building itself a residence.
2. The "I" on a dating profile or résumé. You write, "I am a thoughtful, ambitious person who loves hiking and biographies." This is exactly the case the 論 mentions first: over 色受想行識 (the five aggregates that constitute this being) we erect 自性言論 — "essence-discourses" about who we are. The named "I" is a 假相有法, built across a non-single cluster: the body that ages, the felt tones of recent weeks, the memories of trails and books, the intentions projected forward, the consciousness that will not be the same consciousness five years from now. The name is a guest-perception (客想) added onto the cluster. The *most common mis-mapping*: confusing the guest-perception with the guest's reception desk — believing the named "I" *is* the being, and then suffering when the named "I" is contradicted by the cluster (a body that fails, an ambition that flags, a biography that didn't happen). In 假想 terms, that suffering is the cost of having believed the overlay was the floor.
3. "This is my partner / my child / my parent." Relationships are almost always 依止非一聚 假想物. The named object "my partner" depends on a coordination of behaviors, shared history, mutual expectations, household routines, social recognition by others, future plans, sexual life, financial entanglement, and the way each party speaks of the other in their own internal narration. When the cluster rearranges itself substantially, the name often cannot be sustained — and the work of *renaming* (friend, ex, co-parent, memory) is precisely the 捨舊名、生新名 operation the 論 describes. Where moderns most often slip: treating the name as the *truth* of the relationship and then demanding that the partner actually behave as the name requires. The name becomes a 客法 that demands the cluster conform to it. Inverting the actual order — the cluster first, the name a flexible overlay — is one of the quieter freedoms 假想 analysis offers.
4. A brand you love. "Apple," "Patagonia," "Tesla," "Nespresso," a particular café chain. A brand is a near-perfect 假想物, 依止非一聚: it is assembled from products, design language, advertising, store experience, customer service, the demographics of who else buys it, the public statements of its leaders, the stories told about it, and your own memory of its packaging. Remove the advertising arm and the brand weakens; remove the design language and the brand weakens; remove the store experience and a great deal of the 假想 evaporates. Yet the 實物有 moments — the visible product, the tactile phone, the taste of the espresso — are doing their bare work across many other assemblies (other phones, other espressos). The *most common mis-mapping* is to feel that you have a *relationship with the brand* — that the brand is the guest rather than the guest. Some of the most expensive consumer suffering comes precisely from confusing 客想 with the host.
5. The LLM that hallucinates. A contemporary resonance. When a large language model fabricates a citation, a date, the name of an author, it is doing something functionally similar to 假想: it takes real underlying tokens (a kind of 實物有, perceptually present in the training distribution) and assembles them into a named object (the citation, the date, the person) that may have no actual corresponding 聚 in the world. The result is exactly what the 論 would call a 假相有法: the named object holds its shape within the model's output, but does not correspond to any aggregate of real citations in reality. This is a useful analogy because it shows 假想 analysis doing real work in the contemporary epistemic crisis: many of the things circulating on screens are 假相有法 all the way down — names without clusters. *But* — and the limit must be marked — the LLM is not engaged in 戲論之想 in the Yogācāra sense, because it has no 識, no 想, no 思 as the 論 understands them. The analogy illuminates the *structure of the output*; it does not establish that AI consciousness, agency, or liberation is at stake. Treating the analogy as an identity is the kind of mis-mapping the 論 itself warns about.
6. The news cycle. A "story" is a 假施設 over multiple aggregates: reported events, photographs, named sources, editorial framing, the platform on which it appears, the reader's prior associations, the headline. Same event, different clusters of these conditions → different named "stories" that cannot both fully correspond to the underlying 聚. The 論's move is not "all news is fake"; it is "the named news is a guest-perception on top of an underlying assembly that is itself shifting and partial, and the practitioner's task is to 数数除遣 — repeatedly dismantle — the proliferation while still 唯取其義而轉, taking only the meaning and acting on it without being owned by the overlay."
7. A diagnosis. "I am depressed," "I am ADHD," "I am high-functioning anxious." A diagnostic label is a 客法 over a non-single cluster: symptoms, history, neurochemistry, social context, the clinician's framing, the patient's own narrative, the language of the surrounding culture. The label can be *useful* (依止一聚 of clinical decision-making), and the label can also become a self-imposed residence (戲論之想 building itself a home in the mind). The 唯客想 practice the 論 recommends — recognizing the label as a guest-perception — is *not* a recommendation to drop a clinically helpful category. It is a recommendation to remember that the named object is sustained by a cluster, that the cluster can be worked with, and that the guest-perception need not own the host.
In each case, the move is the same and is the classical one: identify the named object, trace its dependence on one or more aggregates, see the asymmetry between the imputed named object and the bare percepts pervading many assemblies, and — in the practice layer — work with that seeing rather than with the overlay.
当代人为什么需要它
Three pressures make 假想 analysis unusually relevant now.
The inflation of identity-talk. Contemporary life is saturated with 言說相狀 directed at the self. Career narratives, dating profiles, bios, pronouns-in-context, political self-identification, lifestyle brands, recovery identities, spiritual identities. Each is a 客法 added onto a non-single cluster, often with strong emotional charge attached to the overlay. The classical analysis gives a precise vocabulary for noticing when the guest-perception has begun to run the household, and a precise practice (唯客想 + 數數除遣) for loosening that grip without collapsing into a cynical "nothing matters" — because the underlying 聚 and the bare percepts remain real and operative the whole time.
The construction of multi-cluster realities. Brands, institutions, platforms, ideologies, and even relationships are increasingly *engineered* as 依止非一聚 假想物 — explicitly designed to assemble many aggregates into a single named object so that the named object can command attention, loyalty, and money. Yogācāra analysis is, in effect, a diagnostic toolkit for the constructed environment. The 論's distinction between 依止一聚 and 依止非一聚 is exactly the distinction a contemporary practitioner needs to keep clear when distinguishing "the cup of tea in front of me" from "the lifestyle identity being sold through the cup."
The epistemic crisis around generated content. When language models, image generators, and recommendation engines can produce 假相有法 at scale — names without stable underlying 聚 — the analysis the 論 has been quietly refining becomes technically necessary. Learning to ask, of any circulating statement, *what aggregate sustains this name, and what range does the name actually cover*, is the contemporary form of asking the question the 論 asked about eye-name and eye-perception. The practice is not to disbelieve every named object, but to remember that 客法 can be added on top of any 緣 (condition) and to verify the cluster rather than worshipping the overlay.
Finally, the 假想 analysis gives modern practitioners something they tend to lack: a non-nihilistic, non-repressive way to loosen the hold of naming without telling themselves that the world is unreal. The 論 never asks the practitioner to deny 香味觸, to deny the cluster, or to deny the value of meaningful action. It asks for 数数除遣 of the 戲論之想, with 無分別、無相之心唯取其義而轉 — repeated dismantling of the proliferative overlay, with a non-discriminating signless mind that operates by meaning. For someone exhausted by identity inflation on one side and nihilistic detachment on the other, this middle path is the practical gift of 假想 analysis.
常见误读与澄清
1. 假想 ≠ imagination. The English word "imagination" drifts toward fantasy, daydreams, things that are not there. 假想 has nothing to do with the absence of a thing. The 論's whole point is that there *is* an underlying 事 or 聚; 假想 is the *superimposition* of a name and concept onto that thing so it can be talked about. To translate 假想 as imagination is to translate 火 as "cool thing" — formally possible, materially false.
2. 假 ≠ false / nonexistent. This is the single most damaging misreading. The 論 explicitly contrasts 假想物 with 實物有 and concludes that 實物有 is *what the 假想物 is built out of*. The 假 is in the naming-and-assembling, not in the underlying moments. A house named 宅舍 is not an illusion; it is a name overlaid on an aggregate of form/odor/taste/touch coordinated by intention. The dismantling of 客想 does not demolish the house; it loosens the grip of the overlay that had begun to mistake itself for the foundation.
3. 假想 ≠ 想蕴 (saṃjñā-skandha). 想蕴 is the aggregate of perception as one of the five a clinging being grasps at. 假想 is a *specific operation* within the activity of 想: namely, the operation by which 想 takes up 言說相狀 and lays a name onto a 事 or 聚 so 言說 can run on it. Conflating the two flattens the very distinction the 論 is trying to make. The 《心經》 line 「照見五蕴皆空」 is about the five aggregates as such, including 想蕴; the 假想 analysis is about *one specific mechanism by which 想蕴 becomes a vehicle for 戲論*. They are related but not interchangeable.
4. 假想 is not only a linguistic matter. Because the 論 frames 假想 through 名 and 言說, modern readers sometimes conclude this is "just" a philosophy of language. It is not. 想 itself takes 言說相狀 as its object — meaning perception is already structured by the linguistic shape before any speech is uttered. The dismantling in practice (數數除遣, 無分別, 無相) operates on the perceptual-cognitive layer, not on the words. Words are the most visible symptom of a perception that has already been told what it sounds like.
5. Dismantling 假想 is not denial of meaning. The 論's practice is 数数除遣一切戏论之想, *with* 無分別、無相之心唯取其義而轉 — repeatedly dismantling the discursive overlay while *still taking the meaning and acting on it*. This is not quietism and not nihilism. A practitioner who has seen through the 客法 of "partner" can still love, still commit, still show up; the difference is that they are no longer owned by the overlay. The same holds for vocation, citizenship, faith, and care. The 論 does not ask the practitioner to abandon any of these; it asks the practitioner to stop being *ruled* by the named overlay as if it had independent self-nature.
6. 假想物's two subtypes (依止一聚 / 依止非一聚) must be kept distinct. Many practical errors come from treating all named objects as if they were single-cluster. A cup of water can sometimes be calmly named and engaged with; an entire "career identity" assembled from training, networks, public reputation, lifestyle, future projections, and self-narrative is *structurally* a different kind of object and demands a structurally different kind of loosening. Collapsing both into one rubric ("it's all just labels") misses the diagnostic precision the 論 is offering.
7. The 客法 argument applies to perception and naming themselves, not only to objects. Because the 論 extends the analysis even to 見聞覺知 — seeing, hearing, sensing, knowing — the practitioner is not left with a final un-analyzed "pure perception" standing outside the analysis. This is the deeper move, and modern readers sometimes flinch at it, treating it as hyper-skepticism. It is not. It is the consistent working-out of the 論's own principle: anything we can linguistically designate as a noun or fix as a perception is, on that very account, a 客想. The practice does not collapse into nothing; it points toward 無分別、無相之心 — a mind that knows without resting on the overlay. What such a mind "is," the 論 leaves for further development; what it is *not* — a series of guest-perceptions mistaking themselves for the household — the analysis makes very clear.
In the end, 假想 is one of the quieter but more technically precise concepts in the 《瑜伽師地論》, and one of the more practically load-bearing. It diagnoses how names get laid onto things, how those names build the things we identify with, and how the practice of repeatedly dismantling those names — without denying the underlying moments they were laid on — opens the way to a mind that 唯取其義而轉. For a contemporary reader being addressed, every moment, by engineered multi-cluster names demanding allegiance, that diagnosis is not antiquarian. It is, in the precise Yogācāra sense, work that needs doing.