What this concept is really saying
The setting: a court of words
In the Indian Buddhist world, formal debate (論議, *vāda* / *prakaraṇa*) was not casual conversation. It was a disciplined practice with rules — closer in spirit to a courtroom, a scholarly disputation, or a Socratic examination than to a Twitter thread. Both sides followed procedures for laying down a proposition (宗, *pratijñā*), supplying reasons (因, *hetu*), and offering examples (喻, *dr̥ṣṭānta*). The audience and judges assessed not only whether the content was correct but also *how* it was argued.
This is the world in which the Yogācāra (瑜伽行派) treatises — primarily the *Yogācārabhūmi-śāstra* (《瑜伽師地論》, T1579) and Asa's *Vyākhyāyukti* tradition culminating in the *Saṃdhinirmocana* and especially *Ārya-saṃdhi-nirmocana* family — locate the concept of 言過 (vāg-doṣa), literally "faults of speech." When the *proponent* (立論者) commits one of nine specific faults in how they argue, they "fall into the defeat position" (墮在負處). This is a *technical defeat* in debate — distinct from being intellectually refuted on the merits.
To grasp why this matters, we need to see the three-way taxonomy the texts give us of *how a debater loses* (論墮負):
1. 捨言 — *abandoning one's thesis*. The proponent, sensing trouble, retreats with elaborate verbal niceties and effectively surrenders their position. This is a *voluntary* withdrawal. 2. 言屈 — *being broken in argument*. Having been refuted, the proponent tries to escape by switching the topic, citing new authorities, dragging in external matters, becoming angry and abusive, or simply falling silent. These are *evasive* maneuvers after defeat has already occurred. 3. 言過 — *faults of speech*. Here the defeat lies in the *form* of one's argumentation itself. The content may even be correct, but the speech commits one of nine identifiable faults, and the audience/judges can rule the proponent has lost on that ground alone.
The distinction is conceptually precise. 捨言 and 言屈 are about *position* and *retreat*. 言過 is about the *quality of the speech-act itself* — whether the proponent, in the very act of arguing, has degraded the argument. The Yogācāra masters evidently believed that bad speech is not a cosmetic problem; in a formal inquiry into truth, it is itself a kind of loss.
The nine faults, examined one by one
Let me walk through each carefully, preserving the original taxonomy rather than collapsing them into a generic "bad communication."
1. 雜亂 (vyaṃsana / "mixing") — Departing from the topic under debate and inserting other matters. The classic image is of someone who, asked about impermanence, starts talking about their uncle's health. The fault is *topical integrity*: the speech stops being about *this* question and becomes about whatever else the speaker happens to be carrying around in their head.
2. 麁獷 (audārya / "roughness") — Speech born of anger-driven agitation (掉舉, *auddhatya*) or restless haste. This is the fault of *emotional contamination* of argument. The debater is technically still on-topic, but the tone — the flaring temper, the impatient rushing — has taken over. The texts use two words close in meaning: 憤怒 (anger flaring up) and 躁急 (hot haste); both are forms of 掉舉 (restlessness/agitation), a classical member of the unwholesome mental factors.
3. 不辯了 (adhiṣṭhāna-daurbhāva / "failure to clarify") — The dharma or meaning articulated cannot be understood by the audience or the opponent. This is not a charge that the speech is *wrong*; it is a charge that the speech *fails to land*. The technical point matters: in formal debate, communication is part of the contract. If your audience cannot follow you, you have not actually argued.
4. 無限量 (aparimāṇa / "without measure") — Words about the meaning become either bloated (repetitive, excessive) or starved (insufficient). This is the fault of *proportion*: either drowning the point in verbiage, or saying too little to support it. Both are mis-matches between the volume of speech and the weight of the issue.
These first four, taken together, address four different axes on which speech can fail: topic (雜亂), tone (麁獷), clarity (不辯了), proportion (無限量). Notice how the Yogācāra masters were already mapping the dimensions of speech analytically, much as a modern speech-act theorist might distinguish locutionary content, illocutionary force, perlocutionary uptake, and so on.
5. 非義相應 (a-artha-saṃvādī / "not corresponding to meaning") — This is the largest and most internally complex of the nine, and the one that requires the most careful preservation of its structure. It has *ten* sub-categories, and flattening it into "the speech doesn't make sense" loses what is most interesting about it. The sub-items are:
- 無義 — *meaningless*: speech that has no substantive content. - 違義 — *contrary to meaning*: speech that actively contradicts the meaning at stake. - 損理 — *damaging to reasoning*: speech that undermines the logical structure of the inquiry itself. - 與所成等 — *equivalent to what is to be established*: this is a subtle logical fault — essentially, *petitio principii* or circularity. The proponent offers as proof something that is itself what they were supposed to prove. The argument does no work because the conclusion is smuggled into the premise. - 招集過難 — *inviting further objections*: speech that, instead of settling the matter, provokes additional difficulties. A modern analogue would be a defense lawyer whose closing argument hands the prosecution a fresh line of attack. - 不得義利 — *not obtaining the benefit of meaning*: speech that fails to actually deliver any practical or epistemic gain. It is not wrong, exactly — it is *fruitless*. - 義無次序 — *meaning without proper order*: the substantive content may be correct but is presented in a sequence that prevents understanding. (Note: this overlaps with 第六 項 非時; the texts allow some redundancy at the seams of their classification.) - 義不決定 — *meaning undecided*: the substantive position is itself unstable. (This too overlaps with 第七項 不決定, again showing the classification isn't watertight but is layered.) - 成立能成 (T1579) / 成立已成 (T1602) — *establishing what can / has been established*: a fault where the debater offers as fresh argument something already conceded or already established. The two Yogācāra texts differ slightly in wording but refer to the same sub-item. - 順從不稱理的諸邪惡論 — *following unreasonable heterodox theories*: speech that, rather than arguing from one's own position, simply capitulates to unfounded contrary views.
This fifth category is where the *philosophical* weight of 言過 becomes most visible. It is not just about polish; it is about whether the speech is doing the *epistemic work* that argument is supposed to do. The circularity item (與所成等), the *petitio principii* item, the *inviting objections* item — these are recognizable as formal logical faults. The Yogācāra masters had a fairly sophisticated map of how an argument can fail even while the speaker is technically on-topic.
6. 非時 (also 不以時 in T1579's heading) — "untimely" — What should be said in proper sequence is given in disordered order. This is a fault of *temporal-argumentative order*: e.g., producing one's examples before one's reasons, or rebutting objections that have not yet been raised while skipping the ones that have. (The texts note the term appears in two forms within T1579 itself — 不以時 in the list-heading, 非時 in the explanation. Both refer to the same fault.)
7. 不決定 (aniścaya / "indecisive") — Establishing a position, then destroying it, then re-establishing it, and so on rapidly, so listeners cannot follow what is actually being claimed. This is the fault of *positional instability*: the debater is *flip-flopping* in real time.
8. 不顯了 (adīpa / "not manifest") — This covers three distinct sub-kinds: - Speech that invites mockery — language so awkward or inappropriate that it undermines the speaker. - Answering without having understood the question — leaping to a response before processing what was actually asked. - Mixing elegant and vernacular registers inappropriately — using 典語 (literary, "high" diction) and 俗語 (vernacular, "low" diction) in incompatible alternation within the same argument. This last item is a particularly *linguistic* concern; it suggests the Yogācāra writers cared about stylistic register consistency, not just content.
9. 不相續 (asaṃtati / "discontinuous") — The speech breaks off mid-flow. The argument is simply not completed. (In oral debate this would be losing one's train of thought; in written debate, a manuscript that trails off; in modern online discourse, a reply that never quite lands.)
凡所言論犯此九失，是名言過，墮在負處. As T1579 puts it: any speech committing these nine faults is called 言過, and the speaker falls into the defeat position. T1602's closing formula is parallel: 凡所言論犯此九失，是名言過.
Where this concept sits in the larger system
In the Yogācāra curriculum, debate (論議) is one of the *practices of correct understanding* (如理所引) and is part of a Buddhist training that runs from śīla through samādhi to prajñā. To engage in rigorous debate properly, the texts argue, you need the *skillful means* (upāya) of speech: you must know not only what is true but how to articulate it under adversarial conditions without falling into the very faults that make your speech self-defeating.
It is also worth noticing how 言過 relates to, but is distinct from, the moral category of *wrong speech* (惡口 etc.) familiar from the precepts. Wrong speech in the precepts concerns harm to others (lying, divisive talk, harsh speech, idle chatter). 言過 is *procedural-argumentative*: it concerns the conditions under which one's speech functions as *argument*. You can commit 麁獷 (rough speech) without it being a moral wrong, and you can lie quite smoothly without committing 言過. The two categories cross but do not coincide.
Finally, 言過 sits inside a tradition that included the towering logicians Dignāga (陳那) and Dharmakīrti (法稱), whose treatises on valid cognition and inference (因明, *hetu-vidyā*) shaped how Indian Buddhists understood debate. The Yogācāra classification of 言過 is more procedural and behavioral than Dignāga's tightly formalized *ninefold reasoning* (九句因) or Dharmakīrti's theory of *āptavāda* (authoritative testimony), but it sits in the same intellectual ecology — one that took *how* we argue as part of *whether* we have argued well.
生活里的走查: 言過 in contemporary life
Now we come to the part where this technical taxonomy touches the floor of lived experience. Let me walk through the nine (plus the ten sub-items) by way of scenes that almost everyone reading this will recognize. I want to map the classical categories as carefully here as I did in the scholarly section, because the most common mistake is to treat 言過 as a vague "speak nicely" rule rather than a precise diagnostic.
Imagine a workplace meeting. A team is debating whether to launch a product in spring or push to autumn. The marketing manager opens his remarks by spending ten minutes on the new logo design. This is 雜亂: he has left the topic under debate (launch timing) and introduced another matter. Whether or not the logo matters elsewhere, it is not *this* debate. Modern meeting culture often treats this as merely "rambling," but the Yogācāra diagnosis is sharper: it is a *fault in the speech-act itself* — it stops being the argument it was supposed to be.
Imagine a parent–teen argument about screen time. The parent, increasingly frustrated, starts speaking faster, voice rising, sentences stacking on top of each other. The teen is no longer really listening; the parent is no longer really arguing. This is 麁獷 — anger-flared agitation (�怒發作而掉舉) or hot haste (躁急而掉舉). Both are forms of restless disturbance. The fault here is *emotional*, but it is not a private moral failing only — it is a *failure of the speech as argument*. Once the emotion takes over, the argument loses its argumentative character.
Imagine a strategy presentation where a consultant uses jargon like "leveraging cross-functional synergy to drive stakeholder alignment." Half the room nods politely; the other half is lost. The consultant is committing 不辯了 — the meaning is not actually being conveyed. Notice that the *meaning* of the speech might be sound, but the speech has not *landed*. The Yogācāra diagnosis is that the speech is *non-functioning as communication*, and that is itself a fault.
Imagine a long email thread about a single small decision. Some participants write ten paragraphs; some respond with a single sentence that omits three crucial points. Both are 無限量 — *without measure*. One is excessive (重複繁多); one is deficient (有所減少). The fault is the mis-calibration between the *weight of the issue* and the *quantity of words*.
Now consider a public policy debate where a politician is asked whether a particular policy will work. They begin by attacking their opponent's character. This is 非義相應 with several of its sub-items active at once: it is 無義 (no substantive content on the question), it is 違義 (contrary to the meaning of the policy question), and it may well be 招集過難 (inviting further objections — once you've gone personal, the opponent is free to return the favor). Or consider a pundit who argues that we should support X because X is good. 與所成等 — the supposed proof is identical to the conclusion. The argument performs no logical work; it merely restates.
Consider a long-form essay or podcast that begins with examples, then offers the thesis, then returns to history, then gives the thesis again, then offers reasons. The order is scrambled. The reader cannot reconstruct the argument because the *sequence* of presentation is wrong. This is 非時 (untimely) — sometimes called "不以時" in the list-heading of T1579. The content may be excellent; the *order* destroys it.
Consider a public figure who, in the same press conference, says "we will do X," then "we cannot do X," then "we are committed to X but on our terms," then "X is impossible." This is 不決定 — building and destroying, building and destroying, with such rapid switching that no listener can locate the actual position. The fault is *positional instability*: the speech never settles into a claim that can be evaluated.
Consider a colleague's reply in a meeting — they answer a question no one asked, use one piece of vocabulary in formal register, then switch to slang mid-sentence, then make a comment that the room snickers at. Three sub-kinds of 不顯了 in one speech: language that invites mockery (譏弄), answering without grasping the question (未領會問題便作答), and mixing elegant/vernacular registers inappropriately (典語與俗語前後交替混用).
Consider a Slack thread where someone begins to articulate a careful concern, gets three sentences in, then — *...*. They never finish. The next message is unrelated. The argument simply *stops*. This is 不相續 — discontinuity. The speech has broken off mid-flow; the argument is never completed.
Where the大众most often mis-map
The most common contemporary misreading is to fold 言過 into "be polite" or "be respectful" or "don't hurt people's feelings." That is a moral category. 言過 is not that. A debater can be courteous and still commit 言過 if they wander off topic, answer questions that weren't asked, or build and destroy their position five times in one paragraph. Conversely, a debater can be harsh and yet not commit 麁獷 in the technical sense if their words are direct, controlled, and on-topic.
The second common misreading is to treat 言過 as a checklist of "speech etiquette tips." It is not etiquette. It is a *diagnosis of when speech stops functioning as argument*. The reason this matters is that the *consequence* of committing 言過 in the classical setting is severe — you are judged to have lost the debate (墮在負處). The Yogācāra masters believed that if your speech is non-functioning as argument, the debate is over, *regardless of whether your content was right*. That is a strong claim, and a useful one to sit with.
The third misreading is to ignore the heaviest category, 非義相應, because it requires real attention to follow its ten sub-items. A reader who skims will hear "not corresponding to meaning" and miss that this is where the *logic* of 言過 actually lives — circularity (與所成等), inviting further objections (招集過難), following unreasonable heterodox views (順從不稱理的諸邪惡論). If you want to use 言過 to improve your own thinking, this is the section to dwell on.
Why contemporary people need this
We are living through what might be called the age of infinite speech. More words are produced, exchanged, and broadcast every day than in any previous century by orders of magnitude. Social media, group chats, comment sections, podcasts, AI-generated text, hot takes, threads, sub-threads, replies to replies — speech has become nearly frictionless. And precisely *because* speech is so abundant, the question of whether speech is *functioning as argument* has become more important, not less.
The Yogācāra diagnosis is striking: in any inquiry, what matters is not just whether the words are produced, but whether they are doing the *work* of inquiry. 九過 (the nine faults) is essentially a diagnostic for *speech that has stopped working as inquiry*. If you are trying to figure something out — with a partner, in a team, in a public discussion, even privately as you write out your own thoughts — you can run your own speech against these nine categories and find the fault.
A few specific reasons contemporary people might find 言過 useful:
1. To diagnose online discourse. Much of what passes for argument online is, by this classical yardstick, not argument at all. It is 雜亂 (off-topic), 麁獷 (anger-driven), 不辯了 (incomprehensible to half the audience), 無限量 (repetitive talking points or cryptic two-word replies), 非義相應 with multiple sub-items active at once, 不決定 (flip-flopping), 不顯了 (mixing registers, inviting mockery), or 不相續 (threads that trail off into nothing). The Yogācāra list is, unintentionally, almost a perfect taxonomy of social-media pathology.
2. To check one's own thinking. When you write a long email or prepare a presentation, you can self-audit: am I answering the question that was asked (不辯了, 非時)? Is my position stable or am I flip-flopping (不決定)? Is my argument circular (與所成等)? Am I inviting objections I haven't thought through (招集過難)? This is a tool for *epistemic hygiene*.
3. To distinguish losing a debate from losing one's argument. The three-way classification (捨言 / 言屈 / 言過) reminds us that *defeat* has different causes. Sometimes you lose because you abandon your thesis. Sometimes because you get cornered and try to escape. Sometimes because your speech itself is non-functioning. These are different lessons, and conflating them keeps you from learning the right one.
4. To resist the false equivalence between AI fluency and good argument. Large language models can produce speech that is fluent, on-topic, polite, and continuous — and yet *commit 言過 at every level*. Their speech may lack the *topical integrity* required by 雜亂, the *epistemic work* required by 非義相應 (the circularity sub-items are particularly easy for pattern-matchers to slip into), the *positional stability* required by 不決定, and the *completeness* required by 不相續. A useful reminder: the surface quality of speech is not its argumentative quality. The Yogācāra masters would not have been impressed by smooth, continuous, on-topic AI-generated text that does no real epistemic work.
I should add the standard caveat that comparisons between classical Buddhist analysis and contemporary cognitive science or AI are *analogies*, not equivalences. We cannot claim that neuroscience has "verified" 言過, nor that AI exhibits 言過 "in the same way" humans do. What we can say is that the *structure* of the classical analysis illuminates features of contemporary speech that we might otherwise miss.
Common misreadings and clarifications
Misreading 1: 言過 is just "be polite." As noted above, this conflates the procedural/argumentative category with the moral category of wrong speech (惡口). The fix: 麁獷 (roughness) is the *only* sub-category of 言過 that overlaps with moral wrong speech, and even there it is about *agitation in argument*, not harm done to others.
Misreading 2: The nine faults are just "bad communication tips." The structure of 非義相應 with its ten sub-items shows that the list is not generic. It is specifically about *the logic of speech-as-inquiry*. Treating it as tips flattens the most philosophically interesting part.
Misreading 3: 言過 is about the opponent's feelings. It is not. A debater can commit 言過 against an opponent who is happy, attentive, and comfortable. The fault lies in the *speech itself*, not in its emotional reception. The audience-judge (聽眾/裁判) is part of the picture in 不辯了 and 不顯了, but only insofar as their non-understanding is *evidence* of the fault, not as the locus of harm.
Misreading 4: All nine are equally serious. The texts do not say so explicitly, but it is reasonable to read 非義相應 as the heaviest category, because its ten sub-items reach into the *epistemic content* of the argument. 雜亂 is a more local topical fault; 非義相應 can be a deep logical one. Different commentators in the Yogācāra tradition may rank them differently, but the structure of the list invites this reading.
Misreading 5: 言過 is something only monks and scholars need. The classical setting is formal Buddhist debate, but the underlying activity — *trying to figure things out together through speech* — is universal. Family conversations, board meetings, classroom discussions, even internal deliberation all run into 言過. The Yogācāra list works as a diagnostic for *any* collaborative inquiry.
Misreading 6: Once you avoid the nine, your speech is automatically good. The list is a list of *faults* — it is a negative diagnostic. Avoiding 言過 is necessary but not sufficient. The positive counterpart — what good speech-as-argument looks like in this tradition — is something the Yogācāra texts address elsewhere, in their treatment of *correct* debate procedure (如法論議) and in the broader cultivation of right speech, right intention, and wise attention (yoniśo manaskāra). 言過 is one slice of that larger picture.
A closing reflection. What strikes me most about this classical concept is the *severity* of its judgment: *墮在負處* — you fall into the defeat position. The Yogācāra masters are saying that speech is not decorative around argument; speech *is* argument's vehicle, and if the vehicle breaks down, the journey cannot continue. In a period when we are drowning in speech, when AI can produce infinite fluent words, when a hot take can be posted in seconds and a careful argument can take hours, the ancient insistence that *how* we speak determines *whether* we have spoken well is unexpectedly modern. Nine specific faults. A precise diagnosis. And, if you take it seriously, a path to cleaner thinking.