What this concept is really about
"Similarity" sounds like a humble word. In everyday speech we say "these two things look similar" and move on. In the Yogācāra (唯識 / Yoga Practice) treatises — specifically the *Yogācārabhūmi-śāstra* (《瑜伽師地論》) and the *Vijñaptimātratā-vṛtti / Saṃdhinirmocana* tradition that crystallized in the *Abhidharma-samuccaya* and *Mahāyāna-sūtrālaṃkāra* lines, with the *Sphuṭārthā* commentary lineage feeding into Asanga's *Abhidharmasamuccaya-bhāṣya* (《顯揚聖教論》) — 相似 is something far more interesting than a casual glance. It names the precise cognitive operation by which the mind takes two things, sees that they approximate each other, and stops short of declaring them identical. Similarity is the gray zone between "same" (同) and "different" (異), and most of our judgments live there.
The classical sources never treat 相似 as a single flat meaning. Two great treatises handle it differently, and the differences themselves are instructive:
**The *Yogācārabhūmi* (《瑜伽師地論》) uses the term in three registers, none of them elaborated into a system:**
1. As a category name in a list of virtuous qualities. We find "no-inverted practice, retreat, advance, resemblance, real virtue, well-tamed beings" (無顛倒加行、退墮、與勝進、相似、實功德、善調伏有情) — resemblance sitting alongside real qualities, not subsumed by them. 2. As "similar dharmas" (相似法) in three concrete contexts: *by the similar dharma of nurturing beings* (由養育眾生相似法故), *by the similar dharma of concealing* (由藏竄相似法故), and *by the similar dharma of peril* (由險難相似法故). Each is a real-life situation where something functions *as if* it were something else — a near-equivalent, a partial analog, a stand-in. 3. As "similar turn" (似轉) — a technical term in scriptural exegesis (釋經分別法), alongside "general meaning," "same-class," and "latter-three-seek." It marks the place where a term in one passage is *almost* but not *quite* the same as the term in another; the commentator must flag this near-equivalence rather than letting the reader assume identity.
**The *Vijñaptimātratā* exposition through the *Abhidharmasamuccaya* / 《顯揚聖教論》 does what the *Yogācārabhūmi* only gestures at: it builds a system.** Similarity here is subsumed under "同類 / same-category" (which the treatise defines as *resemblance on a given aspect*), and it is broken into four kinds:
一、自體相似 — Self-nature resemblance. Two things resemble each other in their form and appearance (相貌更互相似). A mango and a mango-like fruit; a face and a face-print.
二、業相似 — Action/function resemblance. Two things resemble each other in what they do, regardless of what they look like. A spoon, a spork, and a pair of chopsticks may differ in form but converge in the operation of "delivering food to mouth."
三、法相似 — Dharmic-aspect resemblance. This is the deepest one. Different *dharmic characteristics* resemble each other through relational chains. The text gives two illustrative chains: - *Impermanence → Suffering → Non-self → Birth → Aging → Death* — each link "more resembles" the next. - Pairs like *form and formless*, *visible and invisible*, *resistant and non-resistant*, *tainted and untainted*, *conditioned and unconditioned* — vast lists of "more resembling each other" (更互相似) across doctrinal dimensions.
四、因果相似 — Cause-effect resemblance. Causes and effects resemble each other in their "can-produce / is-produced" (能成 / 所成) relationship. The seedling resembles the tree; the practice resembles the liberation.
The text then places 異類 (dissimilarity / different-category) as the mirror image of 相似, structured across the same four dimensions. Similarity and dissimilarity are paired, perspectival judgments — same text, same four frames, just flipped.
The treatise also flags a darker use of 相似: it is one of the conditions that can cause a present-tense perception to be confused (現見錯亂), listed alongside illusion, dream-intoxication, insanity. Here 相似 is not the clean doctrinal category but a cognitive hazard — seeing something as what it merely resembles.
Finally, a verse reminds us: *"Mind-dharmas should not be probed conceptually, because they operate via resemblance-objects"* (心法不應思，相似境轉故). The mind does not deal with reality-as-such; it deals with objects that resemble what reality might be.
So where does this sit in the whole liberation project (解脫道)? In Yogācāra, the whole architecture of awakening is about seeing through the way mind constructs its objects. 相似 sits inside that construction: it is one of the ways the discriminating mind (分別心 / vikalpa) operates — by classifying, comparing, and approximating. Wisdom (般若 / prajñā) does not abolish resemblance; it sees how the mind has been making the comparison, and stops mistaking the comparison for the thing compared. The fourfold system in 《顯揚聖教論》 is, in a sense, a map of *how we mistake resemblance for identity*. The 似轉 term in 《瑜伽師地論》 is a reminder that even within scriptural interpretation, near-equivalence must be flagged, never silently equated.
Life walkthrough — running the concept through your day
Most contemporary writing on Buddhist concepts drifts into vagueness. The classical sources won't let us be vague here, because 相似 is itself a precise tool. Let's run the four types of resemblance through a single day, then add the other registers.
自體相似 (Self-nature resemblance) — appearance-level likeness
You walk into a coffee shop you haven't been to before and for a moment think you're in your regular place. The signage, the counter, the barista's apron — the form matches. Two minutes later you realize the menu is different. Your mind had locked onto resemblance at the self-nature level and skipped the verification step.
Where this gets misapplied: Modern identity politics, branding wars, and aesthetic plagiarism disputes are mostly battles at this level. We see two faces that *resemble* and decide they must be the same kind of thing. We see two websites that *look* alike and assume one copied the other. The classical term 相似 is more cautious: it says resemblance stops at the 相貌 level — it is not a verdict on what the thing *is*, only on what it appears like.
業相似 (Action/function resemblance) — functional likeness
Your phone's "swipe up to dismiss" gesture works the same way in three different apps. Different icons, different colors, different companies — but the operation is shared. You learn one gesture and transfer it to the others without thinking. This is 業相似 at work: not similarity of form but similarity of what the thing does.
Where this gets misapplied: Two relationships can share the same operation pattern (one partner withdraws, the other pursues; both eventually apologize first) without sharing the same emotional content. We mistake 業相似 for identity ("all my relationships are the same") when really only the function shape is shared. Conversely, we can fail to recognize 業相似 when we're focused on surface difference ("we're nothing alike" — but both of you are doing the same dance of avoidance).
法相似 (Dharmic-aspect resemblance) — relational-chain likeness
This is the one that takes the most attention. The classical example chains impermanence → suffering → non-self in sequence: each characteristic *more resembles* the next. Translated into contemporary life, it sounds like:
- Chronic overwork → sleep collapse → irritability → relationship damage → regret → overwork to repair. - Each link resembles the next not because they look alike (they don't — one is a schedule, one is a state, one is a behavior), but because of the doctrinal/structural continuity: impermanence gives rise to suffering, suffering discloses non-self.
The second classical list — *form and formless, visible and invisible, resistant and non-resistant, tainted and untainted, conditioned and unconditioned* — is teaching us that doctrinal categories themselves resemble each other through their relational architecture. "Formless" is defined *in relation to* "form"; one cannot be stated without evoking the other. This is 法相似 at the doctrinal level.
Where this gets misapplied: We often try to compare isolated categories and find no resemblance. "How is anger like generosity? They aren't alike at all!" Right — but the 相似 here is *not* at the self-nature or functional level. It is at the level of how each is related to the whole doctrinal map. Anger and generosity are not self-nature-similar; they may even be action-similar in narrow contexts (both can motivate action); but at the 法 similarity level, each occupies a defined position in a system that makes each more intelligible through the other. Misreading this level as level one or two flattens a rich doctrinal map into a flat list.
因果相似 (Cause-effect resemblance) — productive likeness
You write for an hour every morning. Three months later you have a draft. The morning writing resembles the draft — not in form (one is notes, one is prose) but in the 能成 / 所成 (can-produce / is-produced) relation. Cause resembles effect because the effect is the cause continuing into a new moment.
Where this gets misapplied: We look for cause-effect resemblance when there is only self-nature or function resemblance. Two people both working hard ≠ one causing the other's outcome. Two companies with similar logos ≠ one causing the other's brand. 因果相似 is the strictest of the four — it requires a productive relation, not just a visual or functional echo.
The other registers
- **Three 相似法 from the *Yogācārabhūmi*** — by the similar dharma of *nurturing beings*, by that of *concealing*, by that of *peril* — read like a teaching on stand-ins: a parent who nurtures but is not the child; a hiding place that protects but is not the danger; a near-escape that resembles escape without being one. These are moments in life where something functions as if it were something else, and our practice is to recognize the resemblance without collapsing it into identity. "My partner is like a parent to me" — true at the 相似法 level, false if turned into "my partner is my parent."
- 似轉 (similar turn) in exegesis — the closest contemporary cousin is paraphrase that drifts: a teacher rewords a teaching and the reworded version *almost* matches the original but introduces a subtle shift. 似轉 is the discipline of noticing and flagging the shift rather than letting the paraphrase stand in for the source. This is alive whenever you read a translation of a translation, or hear someone summarize a sutra — the summary may resemble the sutra but be off by a crucial degree.
- 相似 as cognitive distortion — the treatise lists resemblance alongside dream, intoxication, hallucination, and insanity as a condition that can make a present-tense perception wrong. Pareidolia (seeing a face in the clouds), the uncanny valley (a robot almost-but-not-quite human triggering discomfort), a friend's name popping up everywhere after you just heard about their divorce — these are 相似 operating not as a clean doctrinal category but as a hazard. The mind locks onto the partial match and runs with it.
- 相似境 (resemblance-object) — *"Mind-dharmas should not be probed conceptually, because they operate via resemblance-objects."* The mind never grasps reality-as-such; it grasps objects that resemble what reality might be. The icon on your phone resembles the function of "home"; the ringtone resembles the event of a message; the emoji resembles the emotion. Your mental life is built from resemblance-objects, not from reality-bits. This is a deep statement about the constructed nature of cognition, and it sits at the foundation of Yogācāra's whole project.
Why contemporary people actually need this concept
We live in the age of resemblance saturation. AI-generated text resembles human writing. Deepfake video resembles a real person speaking. Algorithmically curated feeds resemble our tastes because they have learned our patterns. Recommendation engines propose "things you might also like" — a textbook 相似境 operation. We are drowning in things that resemble other things without being them.
Several specific reasons this concept is needed now:
1. To stay sane in the similarity-saturated environment. Modern life runs on resemblance without identity — and most of our miseries (scams, parasocial relationships, mislabeled products, copycat controversies) come from collapsing resemblance into identity. The fourfold system in 《顯揚聖教論》 is essentially a diagnostic: *which level of resemblance is actually operating here?* If it's only 自體相似 (looks alike), the practical and ethical conclusions are very different from if it's 因果相似 (one actually produced the other).
2. To think clearly about pattern recognition in AI. Modern AI is built on resemblance-matching at the 業 and 法 levels — function resemblance, characteristic resemblance. The verse *"mind-dharmas should not be probed conceptually, because they operate via resemblance-objects"* has a strange new ring when we watch large language models operate on tokens that resemble human language without being human thought. The classical insight — that operating on resemblance-objects is real cognition but not contact with ultimate reality — maps uncannily well, while still differing in crucial ways (LLMs do not have the subjective stream that the Yogācāra mind has, so any analogy is partial, not proof).
3. To handle the cognitive hazards the treatise warns about. The list of similarity-induced distortions (夢醉顛狂 / dream, intoxication, madness) now has new members: algorithmic filter bubbles, doomscrolling-induced numbness, the pareidolia of conspiracy-thinking (every event resembles the pattern). The teaching that resemblance can become a condition for confused seeing is one of the most practically useful warnings in the classical material.
4. To do the inner work of "is this the same as what I went through before?" Much psychological suffering comes from misapplied 相似 — treating a current event as 因果 similar to a past trauma (it resembles, but the causal chain is different); treating a current partner as 業 similar to an ex (the behavior pattern matches, the person doesn't); treating a current failure as 自體 similar to an old identity ("I am a failure" because two moments look alike).
5. To read scripture and tradition with proper 似轉 discipline. Every time we pass a teaching through translation, paraphrase, or interpretation, we are doing 似轉 — turning the meaning across a near-but-not-identical gap. The classical term is a built-in warning against glib equivalence.
Common misreadings and clarifications
Misreading 1: "Similarity means 'almost the same,' so the difference doesn't matter." This is the opposite of what the texts teach. Similarity is precisely the category for when the difference matters but has not been fully tracked. The whole reason the treatise breaks it into four types is so the difference can be tracked at the right level. Collapsing similarity back into identity is exactly the cognitive error the concept is designed to keep visible.
Misreading 2: "Similarity is just a visual thing." This comes from collapsing the four types into 自體相似 (appearance resemblance) only. The classical material is explicit that resemblance can be at the level of form (相貌), function (業用), doctrinal-characteristic-relation (法門差別), or productive causality (能成所成). Conflating all resemblance with the visual level strips the concept of its diagnostic power.
Misreading 3: "If two things are similar, they're in the same category." This reverses the technical definition. 《顯揚聖教論》 defines 同類 (same-category) *as* similarity-on-a-given-aspect — but 同類 is not the precondition for similarity; it is the name for when similarity has been established on a given aspect. Similarity is the underlying cognitive operation; 同類 is the classificatory verdict that follows. The treatise also defines 異類 as the inverse — not "different things" in general, but "dissimilar on a given aspect." Both 相似 and 異類 are perspectival: change the aspect, change the verdict.
Misreading 4: "Similarity is bad — it leads to confusion." The cognitive-hazard use of 相似 is one of the use-cases, not its essence. Resemblance is also a legitimate tool: it is how we learn (the new resembles the known), how we transmit teaching (the example resembles the principle), how scripture works (the parable resembles the truth it points toward). The 似轉 term in exegesis is precisely a *disciplined* use of resemblance in scriptural interpretation. The concept doesn't condemn resemblance; it asks us to be honest about which kind is operating.
Misreading 5: "The four types are rigid categories." They are diagnostic frames, not airtight boxes. A single situation can exhibit resemblance at multiple levels (the coffee shop resembles your regular one in form, in function, and in the emotional relief it offers), and the classical text explicitly uses phrases like 展轉相似 (resemblance across links) and 更互相似 (mutual resemblance) to indicate that resemblance can chain. The discipline is to ask which level is doing the work in a given judgment, not to find the one true box.
Misreading 6: "This is just Yogācāra technical vocabulary, irrelevant to practice." The verse *"mind-dharmas should not be probed conceptually, because they operate via resemblance-objects"* is itself a practice instruction. It is telling the meditator: when you turn the mind inward and try to grasp it directly, you will not find a thing — you will find resemblance-objects, because that's the medium the mind uses. This is foundational for the Yogācāra practice of seeing how the three natures (三性) operate: the imagined nature (遍計所執性) is built largely out of resemblance-collapsed-into-identity errors; the dependent nature (依他起性) operates through resemblance-objects; the perfected nature (圓成實性) is what is seen when resemblance is recognized as resemblance.
Misreading 7: "If similarity is perspectival, then anything goes." The fact that 相似 and 異類 are perspectival does not mean they are arbitrary. The text grounds them in four structural dimensions (自體, 業, 法, 因果) that are not chosen arbitrarily by the observer — they are features of how phenomena actually relate. The perspectival element is which aspect you are examining, not the verdict itself. This is a subtle but important difference: the verdict of similarity or dissimilarity follows the structure of the phenomena on the aspect chosen; it is not free-floating opinion.
A closing thought. The Yogācāra teaching on 相似 is, in the end, an instruction in honest seeing. Most of our errors — personal, social, political, contemplative — happen because we treat resemblance as identity, or identity as resemblance, or because we fail to track which level of resemblance is actually operating. The classical sources give us not a theory to believe but a discipline to practice: when you notice two things seem alike, ask — *alike in form? alike in function? alike in their doctrinal position? alike as cause and effect?* — and let the answer stay humble. Similarity is a real cognitive operation. It is not the same as the thing compared. And the mind that knows the difference is closer to the kind of seeing the whole path is asking for.