What this concept is actually about
If you walk into a contemporary bookstore and look for "類," you would find it on the shelf of everyday Chinese as "type," "kind," "category." That's a perfectly fine reading — but it's not the reading the *Abhidharmakośa* (《阿毘達磨俱舍論》) is doing. In this dense treatise, *lèi* is not a noun that stands on its own. The text doesn't posit an abstract "class" the way a logician might. Instead, *lèi* works almost as a verb-like connector: it marks where similarities hold, where they don't, and what counts as a legitimate *belonging* relation between dharmas (法). To get it, you have to follow the words it shows up inside.
The single most important place it shows up is the compound 同類因 (tónglèi yīn) — "cause of the same kind." The text defines it crisply: "相似法與相似法為同類因" — *similar dharmas serve as the cause-of-the-same-kind to other similar dharmas*. The example given is the five aggregates (五蘊, *pañca-skandha*) of goodness, the five aggregates of defilement, and the five aggregates of the indeterminate (無記), each one looking at its own kind across time. Note the term used here: 五蘊, not 五陰. The *Kośa* uses the "accumulation" rendering, which leaves a faint philosophical footprint — *skandha* as something that accumulates, not as a "shadow" or "obstruction" — and that footnote matters in this context because the whole point of *lèi* is about how things accumulate and stand together.
So far, this could pass for a polite philosophical truism — "similar things go together." But the *Kośa* then adds three sharp restrictions that turn a soft truism into a hard-edged technical concept. Each restriction is a separate move, and each must be kept distinct.
The first restriction is by 部 (bù), "department." The text says: "自部自地唯與自部自地為因" — *only within one's own department and one's own ground does the cause-of-the-same-kind relation hold*. Five departments are enumerated: 見苦所斷 (cut-off-by-seeing-suffering), 見集所斷 (cut-off-by-seeing-origination), 見滅所斷 (cut-off-by-seeing-cessation), 見道所斷 (cut-off-by-seeing-the-path), and 修所斷 (cut-off-by-cultivation). Each department only causes effects within its own department. If you collapse them into a single "type," you've lost the whole point. Two things can be similar in nature and still not be of the same kind, because they belong to different *stages of overcoming*.
The second restriction is by 地 (dì), "ground." Nine grounds are listed: one for the desire realm (欲界), four for the meditation absorptions (四靜慮), and four for the formless absorptions (四無色). Each ground only causes effects within its own ground. The same kind of mental state in the desire realm and in the first dhyāna may look similar, but cross-ground causation of the same-kind type does not occur. The *lèi* is not just a matter of flavor; it is pinned to a specific cosmological-existential location.
The third restriction is by 前後生 (qián hòu shēng), "prior and subsequent arising." The text borrows from the *Jñānaprasthāna* (《發智論》): "前生善根與後生自界善根及彼相應法為同類因" — *prior-arising wholesome roots serve as cause-of-the-same-kind to later-arising wholesome roots of the same realm and their associated factors*. This is the moment where *lèi* stops being a static classification and becomes a causal sequence. The "same kind" here is not just "alike in property"; it is "alike in property, and the earlier one is feeding the later one." Similarity plus temporal precedence plus genuine causal efficacy.
Outside the *tónglèi yīn* framework, the *Kośa* extends *lèi* in several secondary directions, and these directions are not interchangeable:
- 自類 (zìlèi) — "one's own kind." The text illustrates this with one body going through ten stages (from 羯剌藍, *kalala*, the embryonic drop, onward) where each stage serves as cause-of-the-same-kind to the next within itself; *and* with the same ten stages of *another* body, where each stage can serve as cause-of-the-same-kind to all ten of that other body. Wheat and rice are likewise treated by "自類自類思擇." So *zìlèi* can be a robust within-one-stream identity (one body ten stages) or an across-streams species identity (wheat to wheat), and in both cases it must still satisfy the causal conditions.
- 類 as a numerical marker of classification. The text uses phrases like "得相通三類" (three classes of attainment and so on), "此中有四種，類相位待異" (four kinds, differing in class, phase, and dependence), and "由結等差別，復說有五種" (five kinds, by the difference of fetters and so on). Here *lèi* functions as a generic count-word for a kind carved out by some specific feature — attainment, sign, phase, dependence, fetter. The text itself warns against merging these: the bases of division differ, so the contents of each "kind" differ.
- 法類 (fǎlèi) and 二類 (èrlèi), the technical compounds. "法類由境別，立苦等四名" — *dharma-class and class-of-dharma are distinguished by their objects-of-cognition, and are given the four names of suffering, origination, cessation, and path*. This associates *lèi* with the Four Noble Truths and with the pair 盡智 (exhaustion-knowledge) and 無生智 (non-arising-knowledge). "二類定順逆" — *two kinds determinately define forward and reverse orders*. Here *lèi* is the unit of a sequence that can be traversed forward or in reverse, such as in the gradual path.
So when you collapse everything and say *lèi* just means "category," you have lost the three restrictions (department, ground, temporal precedence), you have lost the causal structure, and you have flattened half a dozen distinct specialized uses into one fuzzy word. The modern reader who keeps this in mind — that *lèi* is a *constrained* similarity with a *causal* backbone — is already holding the right end of the stick.
Life walkthrough: A scene most people will recognize
Let me walk this through with a scenario familiar to many: a habit of getting anxious on Sunday evenings. We'll trace it strictly along the lines the *Kośa* lays out, and along the way point out where most people instinctively map the word wrong.
First, identify the similarity. You notice that every Sunday evening, around 6 or 7 p.m., a familiar low-grade dread shows up. The texture is similar each time: tightness in the chest, a particular flavor of self-critical thinking, a sudden urge to refresh work email. These are the *similar dharmas* — the same anxiety flavor across nights. So far the *Kośa*'s "相似法" criterion is met. This is the step where most people stop, and conclude they have "a Sunday-evening anxiety type." But the *Kośa* has not stopped.
Second, apply the department (部) check. The five departments are not on the surface of experience, but they are doing real work. The Sunday dread is not identical with the sharp, in-the-moment panic you feel when your boss calls you into a meeting. The latter arises *with* a present object (the meeting), the former arises *in anticipation of* an upcoming object (the week). They are similar in feel, but they belong to different "departments" of how the mind is responding to its objects — one is reacting to a present cause, the other is reacting to a future projection. The *Kośa* would say: these are not the same kind, despite surface similarity. So if you treat them as one "type of anxiety," you have already mistaken surface similarity for *lèi*.
Third, apply the ground (地) check. The same person, on a Saturday morning during a hike, when stress is low and the body is rested, may experience a *flavor-similar* discomfort — a tightness in the chest about the upcoming week. But the ground has changed. The first dhyāna analogy is rough but serviceable: the same kind of mental formation can only appear as cause-of-the-same-kind within its own ground. When you are rested, the chain of "prior Sunday anxieties feeding subsequent Sunday anxieties" is *grounded* differently than when you are depleted. The *Kośa* insists this matters. The habit chain is not just a thought pattern floating free; it is a chain anchored to a specific existential ground.
Fourth, apply the temporal-precedence (前後生) check. Here is where the most counterintuitive move happens. The "Sunday anxiety" that happened five Sundays ago is not just *similar to* the one this Sunday — it is among the *causal conditions* for it. The *Kośa* would frame last Sunday's anxiety as a *prior-arising* of the same kind as a *later-arising* of the same kind. That is, the habit is not just a resemblance; it is a *lineage*. The five-Sundays-ago version is, in some technical sense, feeding the present one. This is the moment where "類" stops being a label and becomes a *causal channel*.
Fifth, drop in the 自類 (zìlèi) distinction. The line of Sunday anxieties is *one body's* ten stages, in the *Kośa*'s metaphor. Your stream of consciousness has been accumulating these as its own kind. Meanwhile, your friend across the city also has Sunday anxieties, but hers are not the same kind as yours — they form a separate stream, with their own line of stages. The grain-vs-grain analogy: their wheat is wheat, yours is wheat, but the field of causation is shared only within streams, not across. This is why "talking to your friend about it" can help in a generic sense but does not *cause-of-the-same-kind* dissolve your own.
Now the misreadings. The habit-of-mind reader might think: I've been told my anxiety is an "anxious-avoidant type" by a personality test. That flattening — putting every form of relational anxiety into one *lèi* — violates the *Kośa*'s department, ground, and temporal distinctions. The MBTI-style identity is a *similarity statement without causal backbone*. The *Kośa*'s *lèi* is precisely the opposite: similarity + causal lineage + ground + department. The contemporary test says "you are alike in kind." The *Kośa* says "you are alike in kind *and* this is the lineage through which the likeness is being reproduced." The difference is the difference between a photograph and a river.
A second scene, more subtle. You start a meditation practice. For three weeks, sessions are pleasant. For the next two, they are dry and uncomfortable. The *Kośa* would not say these are two unrelated "types" of meditation experience. It would ask: is the pleasant session a *prior-arising* of the same kind *feeding* the dry session? Or are they different departments (different stages of overcoming attachments to sensory pleasure)? Or are they different grounds (one in dhyāna-like absorption, one in the desire realm)? Each reading produces a different understanding of what to do next. The *lèi* framing forces you to ask *which kind of kind* this is. That is the work the concept is doing.
Why contemporary people need this idea
There is a flatness in how we categorize things today. It is the very flatness that ancient Indian analysts were pushing against when they insisted *lèi* had to be carved up by department, ground, and temporal order. Three reasons in particular that the concept matters for a contemporary reader.
First, the deluge of similarity claims. Algorithm-driven recommendation systems, personality typologies, mental-health diagnostic labels, ideological sorting — modern life is saturated with statements of the form "X is like Y." Most of these statements are *similarity without causal structure* and *similarity without boundary conditions*. The *Kośa*'s insistence that similar things are not the same kind unless they share department, ground, and causal lineage is a quiet philosophical rebuke of the contemporary tendency to treat shared surface features as deep identity. It is a tool for resisting premature categorization.
Second, the contemporary neglect of causal lineage as a feature of kinds. When neuroscience talks about neural plasticity, it is talking about how prior firing patterns shape later firing patterns. When cognitive-behavioral work talks about schemas, it is talking about how prior interpretive patterns shape later interpretation. When habit-formation research talks about loops, it is talking about how prior behaviors shape later behaviors. These are all *posterior-arising by prior-arising of the same kind* — in the *Kośa*'s technical sense. The contemporary frameworks often *describe* this without naming it as a kind-belonging relation. The Buddhist framing gives a sharper vocabulary: when something keeps happening, it is not because the world is repeating — it is because a *lèi* lineage is being maintained. That distinction is actionable, because lineages can be interrupted.
Third, the multi-axis nature of belonging. The *Kośa*'s insistence that *lèi* is determined by *both* department and *ground* — not one or the other — is a useful corrective to the modern tendency to classify on a single axis. In real life, two people may be similar in psychology (same "department") but exist in utterly different material conditions (different "ground"). Sorting them into the same kind therefore misclassifies them. Recognizing that *lèi* is multi-axis helps in clinical, social, and political contexts where single-axis classification produces real harm.
It is worth being honest about the limits of the analogy here. Neuroscience and AI recommendation systems are not the same thing as *Kośa*-style dharma analysis. The brain is not the *Abhidharma*'s taxonomy. Algorithms do not have consciousness. The structural resemblance is real at the level of *causal pattern*, but the *Kośa* is operating within a much richer ontological framework that includes karma, rebirth, and the structure of awakening. Anyone who pushes the analogy past the structural level will be overclaiming. The usefulness is in the framing, not in the equivalence.
Common misreadings and clarifications
Misreading 1: "類 just means category." This is the most common flattening. The *Kośa* never uses *lèi* as a bare category. Every use is constrained — by department, by ground, by temporal order, by causal efficacy, or by a specific technical role (法類, 二類). Reading it as plain "category" strips away roughly 80% of the conceptual content.
Misreading 2: "If two things are similar, they are the same kind." This is directly contradicted by the department and ground restrictions. Two mental states can be emotionally similar and still be of *different kinds* because they belong to different departments of cognitive overcoming or different existential grounds. The similarity that counts for *lèi* is *constrained* similarity.
Misreading 3: "Same kind means static identity." The temporal-precedence restriction prevents this. The *lèi* relation is dynamic — it is a lineage in which prior-arising and later-arising are tied together. Treating same-kind as a frozen label (you are an introvert, you are an anxious person) is a structural misreading of the *Kośa*'s framework.
Misreading 4: "自類 and 同類因 mean the same thing." They are related but not identical. *同類因* is the causal technical term, requiring the full set of restrictions. *自類* is the broader identity-claim used in compound expressions (one's own kind, wheat's kind, etc.) that still needs to be backed by causal conditions but is more flexible in deployment. The *Kośa* uses them in different contexts and loads them differently.
Misreading 5: "The 五種, 三類, 四種 listings can be merged into a unified taxonomy." The text explicitly resists this. *得相通三類* is divided by attainment; *由結等差別，復說有五種* is divided by fetters; *此中有四種，類相位待異* is divided by class, phase, and dependence. The bases are different, so the contents are not interchangeable. A seal of "five kinds" in one place is not the same five kinds as in another.
Misreading 6: "法類 just means 'the category of dharmas.'" It is more specific than that. *法類由境別* links *lèi* to the Four Noble Truths and to the pair of exhaustion-knowledge and non-arising-knowledge, distinguished by object of cognition. Reading it as a generic "dharma-category" loses the entire connection to the liberation path.
A final caution. The *Kośa* is operating within a specific Abhidharma tradition that has its own metaphysical commitments. The contemporary reader should not assume that adopting *lèi* as a tool requires adopting the full cosmology of nine grounds, five departments, or the rebirth-implied karmic flow. What survives translation into modern life is the *structure* of the analysis: similarity is bounded; belonging is multi-axis; causal lineage is part of what makes a kind a kind. That structure is portable. The full metaphysical scaffolding is not, and the *Kośa* itself would not have asked the reader to mistake scaffolding for the house.
Standing back. The concept of *lèi* teaches, in its quiet technical way, that to say "this is like that" is only the beginning of inquiry. Inquiry must continue: *like in which way? in which department? on which ground? with what causal history? for which stage of the path?* The word that looked like a simple category turns out, on close reading, to be a small philosophical discipline — a way of refusing to let similarities harden into identities before they have earned the right to.