A Corpus Census of the ‘ti-Construction, Its Ancient Readers, and What It Means for the Meditator
A guest contribution. The thesis examined here was first advanced by the editor of this blog in March 2008 (“iti and sallakkheti”) and April 2009 (“Labels employed for noting in early Pitaka”). The present study subjects that thesis to a complete computational census of the Chaṭṭha Saṅgāyana corpus, the GRETIL Sanskrit collection, and the Chinese Āgamas, and reports its findings.
Abstract
The Pāli canon contains several thousand instruction-formulas in which a short quotation closed by the particle iti fills the object slot of a verb of knowing: rūpaṃ aniccan’ti pajānāti. Standard translations dissolve the quotation into a content-clause — “he knows form as impermanent” — and with it a question disappears: why does the canon so often put a quote where we expect a proposition? It also baffles the reader, when translated this way: “Yes, but HOW does he see it as impermanent” – masking what could perhaps be the meditation instruction itself.
This study reports a complete census of the construction: 7,384 tokens extracted from all 217 files of the Chaṭṭha Saṅgāyana Roman-script corpus, classified by explicit criteria, and tested against the Sanskrit and Chinese parallel traditions. The results are asymmetric. In the canonical layer, label-shaped tokens — short, formulaic, person-preserving quotes drawn from a closed lexicon of roughly sixty forms — outnumber clearly propositional uses nearly five to one; the satipaṭṭhāna refrain equates the bare label with established mindfulness; and the distribution of the formula against the jhāna scheme matches, with precision, the canonical doctrine that thought is speech-formation (vitakka-vicārā vacīsaṅkhāro). At the same time, no early text states that the labels are recited; the Chinese translators rendered the formula as plain knowing; and the same syntactic frame freely carries content no one could “note.” The essay weighs both sides and argues that the thesis survives: the canonical formulas are instructions in deployed vitakka — a fixed, first-person thought placed at the point of contact, self-terminating as concentration deepens — and that modern Burmese noting stands at the end of a documented textual pipeline running from these formulas through the commentarial vocabulary of sallakkhaṇā to the practice halls of the present. The closing sections ask what this recovery means for the meditator and for a living Buddhism.
Introduction: A Sentence Hiding in Plain Sight
Somewhere in the second or first century before the common era, the compiler of the Peṭakopadesa — one of the two ancient hermeneutical manuals the Pāli tradition associates with the lineage of Mahākaccāyana — condensed a teaching he knew from the Saṃyutta collections into nine words:
Samādhiṃ, bhikkhave, bhāvetha, samāhito, bhikkhave, bhikkhu rūpaṃ aniccan’ti pajānāti.[1]
Develop concentration, monks; a concentrated monk knows form: “impermanent.” So the sentence reads if one keeps the quotation marks that the particle ‘ti supplies. Nearly every published translation removes them: “a concentrated monk knows form as impermanent,” or “understands that form is impermanent.” The rendering is not wrong — we shall see below exactly how much grammatical right it has — but it is lossy. It deletes from the English page a feature that is unmistakably present on the Pāli page: the content of the monk’s knowing is given not as a description but as a quotation, one word long, in the form a thought takes when it occurs.
In March 2008 the editor of this blog pointed at this loss and made a claim about it: that formulas of this shape are not statements about meditators but instructions to them — that the canon here tells the practitioner to know form while noting “impermanent” — and that the Burmese practice of mental noting, so often described as a twentieth-century invention of the Mahasi school, is in fact the survival of something the texts have carried openly all along.[2] The claim was made from a handful of examples and a search-window count. It deserved better than that, and this study gives it better: a full census.
The method is stated so that a skeptic can repeat it. Every occurrence of quoted material governed by a verb of perception or training — pajānāti, jānāti, passati, samanupassati, sallakkheti, manasikaroti, paccavekkhati, daṭṭhabbaṃ, sikkhati — was extracted mechanically from the entire Chaṭṭha Saṅgāyana corpus: canon, commentary, subcommentary, and the para-canonical works. Each token was classified by explicit, published criteria; every count in this essay is reproducible from the scripts; every Pali, Sanskrit, and Chinese passage cited was verified in the digital corpora, none quoted from memory.[3] The census was then set against the two great control traditions: the Sanskrit scholastic corpus, and the Chinese Āgama translations made between the second and fifth centuries — the earliest external witnesses to how ancient readers understood the formula.
Two hats are worn throughout, and the reader should know which is which. The thesis is advocated: it holds that these formulas teach noting. The census is neutral: it counts what is there, and it has turned up evidence against the thesis as well as for it, all of which is reported. The conclusion, as will be seen, is not the thesis in its boldest form. It is something more defensible and, I will argue, more interesting.
Part I: The Construction and the Census
The Grammar of iti and the Limits of Grammar
The particle iti (enclitic ti) closes a quotation. It is also, and this must be conceded at the outset, Pāli’s only complementizer: the language has no separate “that”-clause, and whatever a person says, thinks, knows, or believes is given in the direct form, with the person and tense of the original utterance or thought preserved.[4] When the Satipaṭṭhāna Sutta says ‘dīghaṃ assasāmī’ti pajānāti,[5] the first person of assasāmi — “I breathe in” — proves nothing by itself: Pāli could not have expressed “he knows that he is breathing in long” any other way. Every single instance of the construction is therefore ambiguous between quotation-as-label and quotation-as-content. The old dictionary judgment that iti is “on the whole untranslatable” and best rendered by quotation marks or italics is the sound of grammarians declining to decide the question.[6]
This concession defines the method. If grammar cannot decide, distribution must; the question is not what any sentence means but what thousands of them, taken together, are shaped like. And here the census speaks with an asymmetry that no single passage could reveal.
A Census of 7,384 Tokens
Across the 217 files of the corpus the extraction found 7,384 tokens of the construction. Classified by explicit criteria, the canonical layer (2,702 tokens) breaks down as follows: 843 tokens whose quote is first-person (“I breathe in long,” “there is in me sense-desire”); 367 bare-predicate labels of four words or fewer in explicit practice contexts (“impermanent,” “a lustful mind”); 670 of the same shape whose context the classifier would not vouch for; 532 retrospective attainment-reviews (“birth is destroyed”); 20 tokens of the satipaṭṭhāna refrain, discussed below; and — the contrast class on which everything turns — only 255 tokens whose quote is a genuine content-clause or a doctrine-summary of sutta length.[7] Label-shaped quotation outnumbers clearly propositional quotation in the canon by nearly five to one.
The commentarial layer inverts the picture in an instructive way: there the dominant use (3,059 tokens) is the exegetical idiom ‘…’ti daṭṭhabbaṃ, “should be understood as: …” — a scholastic construction absent from the canon. The two layers use the same syntax for different work, and keeping them separate is what makes the canonical figures meaningful.[8]
The Closed Lexicon
The strongest structural finding is this. If ‘X’ti pajānāti meant only “he knows that X,” the X-slot should be an open set, for the set of knowable things is open — and the canon does show us what that open use looks like: quotes of paragraph length, future-tense contents, doctrinal catalogues, each essentially unrepeated.[9] A content-clause construction has no reason to have a closed lexicon. But the practice formulas draw from a repertoire so small that some sixty fixed forms cover the great majority of over a thousand tokens, repeated verbatim across all four Nikāyas.[10]
The families are quickly told, and every meditator will recognize the syllabus. The breath: ‘dīghaṃ assasāmī’ti pajānāti — “long I breathe in” — with its fourteen further training-resolves governed by sikkhati, “he trains: ‘experiencing the whole body I will breathe in’.”[11] The postures: ‘gacchāmī’ti pajānāti — going, he knows: “going.”[12] Feeling: ‘sukhaṃ vedanaṃ vedayāmī’ti pajānāti — “I feel a pleasant feeling.”[13] The mind: sixteen pairs, ‘sarāgaṃ cittan’ti … ‘vimuttaṃ cittan’ti pajānāti — “a mind with lust,” “a freed mind” — the same sixteen labels serving both for one’s own mind in cittānupassanā and, remarkably, for the reading of another’s mind in the telepathy pericope: the label set is portable across objects.[14] Self-observation: ‘atthi me ajjhattaṃ kāmacchando’ti pajānāti — “there is in me sense-desire” — through hindrances and awakening factors.[15] The three marks with passati: aniccan’ti, dukkhan’ti, anattā’ti.[16] The anattā formula: ‘netaṃ mama, nesohamasmi, na meso attā’ti — “not mine, not I, not my self.”[17] The ladder of emptiness in the Cūḷasuññata: at each stage, ‘santamidaṃ atthī’ti pajānāti, “there is [only] this, present.”[18]
Four properties run through every family: the quotes are one to four words; the person of the original experience is preserved (assasāmi, vedayāmi, -osmi, me); the vocabulary is fixed to the word across collections; and the governing verbs are verbs of knowing or training — never verbs of speaking. That profile is the fingerprint of a technical vocabulary, a drilled checklist — not of free propositional report. No single sentence proves the labeling reading. The lexicon does what no sentence can.
One further find deserves record here. The Aṅguttara’s Sandiṭṭhika suttas answer the question “in what way is the Dhamma directly visible?” with precisely this construction, turned into the second person: santaṃ vā ajjhattaṃ manosandosaṃ ‘atthi me ajjhattaṃ manosandoso’ti pajānāsi — “when there is defilement of mind in you, you know: ‘there is defilement of mind in me’.”[19] The formula is not merely an instruction among others; it is offered as the very demonstration that this teaching can be checked. Sandiṭṭhika — visible for oneself — is defined by the label.
The Refrain That Names Itself
Twenty canonical tokens stand apart and were classified separately. At the close of each satipaṭṭhāna exercise the text says:
‘Atthi kāyo’ti vā panassa sati paccupaṭṭhitā hoti, yāvadeva ñāṇamattāya paṭissatimattāya.[20]
“Or else his mindfulness is established: ‘there is body’ — just enough for knowledge, just enough for remembrance.” Here the text itself equates the occurrence of the bare label with the establishment of sati, and doubly restricts it with -matta, “a mere measure.” Whatever theory of the construction one holds, this passage is its limit case: the label is not the content of a knowing about something; the label’s presence is the practice, and the text says so. It is the strongest single canonical passage for the minimal-label reading, and the census found nothing on the propositional side of comparable directness.
Part II: The Mechanism — Thought as Inner Speech
Vitakka-vicārā vacīsaṅkhāro
The thesis in its precise form does not claim that the labels were spoken, nor even whispered inwardly in some loose sense. It claims they are vitakka — applied thought, deployed deliberately — and here the canon supplies its own psychology. Dhammadinnā, praised by the Buddha as foremost among his bhikkhunī disciples in wisdom, explains in the Cūḷavedalla Sutta why thought counts as the speech-formation:
Pubbe kho, āvuso Visākha, vitakketvā vicāretvā pacchā vācaṃ bhindati, tasmā vitakkavicārā vacīsaṅkhāro.[21]
“First one thinks and examines, afterwards one breaks into speech; therefore thinking-and-examining is the speech-formation.” The same doctrine recurs in the Kāmabhū dialogue of the Saṃyutta.[22] Thought, in the canon’s own analysis, is inner speech: the phase of the speech act that has not yet sounded. On this doctrine, an instruction to know the breath under the thought “long” is an instruction in subvocal labeling — no verb of speaking required, because the canonical psychology has already located the speech inside the thought.
And this doctrine generates a prediction. The speech-formation ceases in the second jhāna, where vitakka and vicāra fall silent. If the quotative formulas encode deployed vitakka, then label-instructions must cluster at or below the first jhāna and access-concentration, and must vanish inside the deeper attainments — appearing above that line only in retrospection, after emergence. This is a falsifiable claim about the distribution of a construction across a corpus, and the census tested it.
The Jhāna Boundary
It holds. Canonical label-tokens whose paragraphs contain jhāna-attainment formulas number one or two per jhāna, and each is insight practiced upon the attainment from outside it. The 532 review-tokens — “birth is destroyed,” “there is an escape beyond this”[23] — sit after emergence by definition. The apparent counterexamples dissolve on inspection: ‘upekkhakosmī’ti pajānāti, “he knows: ‘I am equanimous’,” occurs in indriya-observation, not inside a jhāna formula;[24] the third jhāna’s own internal quote — ‘upekkhako satimā sukhavihārī’ti — is governed by ācikkhanti: it is what the noble ones declare about the attainer, reported speech, not a meditator’s label.[25] One honest edge remains and is flagged rather than explained away: the formless attainments embed their content-quote in the attainment formula itself — ‘ananto ākāso’ti, “unbounded space” — and the census records thirty-one label-tokens in such contexts, which a defender will read as the saññā-content of the attainment and a skeptic as a breach of the boundary.[26]
The manual stratum then states the self-termination in so many words. The Visuddhimagga’s counting doctrine instructs the beginner to count the breaths with quoted number-labels — ‘ekaṃ ekan’ti … ‘dasa dasā’ti — and fixes the terminus: counting continues yāva vinā gaṇanāya assāsapassāsārammaṇe sati santiṭṭhati — “until mindfulness stays on the breath-object without counting” — for counting exists satisaṇṭhāpanatthaṃ yeva, solely to establish mindfulness.[27] The label is scaffolding, and the tradition knew it comes down.
Nor is the boundary a Pāli idiosyncrasy. The Saṃyukta-āgama, in Guṇabhadra’s fifth-century Chinese, presents the whole doctrine as a ladder of progressive silences: 初禪正受時，言語寂滅；第二禪正受時，覺觀寂滅 — “on attaining the first dhyāna, speech falls silent; on attaining the second dhyāna, vitarka and vicāra fall silent.”[28] The Madhyama-āgama calls vitakka-vicāra the thorn of the second jhāna — 以覺觀為刺 — and the Aṅguttara says the same in Pāli: dutiyassa jhānassa vitakkavicārā kaṇṭako.[29] Across three transmission lines the architecture is identical: discursive labeling is the coarse, useful, self-limiting instrument of the lower floors.
The School That Said It Aloud
That the question “does the label break into speech?” is no modern imposition, the Kathāvatthu proves. Its Vacībhedakathā records a debate with schools — the commentary names the Pubbaseliyas — who held that for one attaining first jhāna at the moment of the path, dukkhanti vācā bhijjati: the utterance “suffering!” literally breaks forth. The refutation proceeds by cornering the opponent: “Dukkhanti jānanto dukkhanti vācaṃ bhāsatī”ti? Āmantā. — “Does one knowing ‘suffering’ utter the speech ‘suffering’? — Yes.”[30] The Theravādin answers no — and note what the dispute concedes on both sides: the opponents place the outbreak of speech precisely and only in the first jhāna, where vitakka still operates; no one imagines speech above it. An early Buddhist school, in other words, read the label-formula so literally as verbal that it had the meditator speak; the mainstream held the label to be vitakka and therefore silent. The position defended in this essay is, quite exactly, the mainstream’s.
The Scholastic Convergence
What the Pāli canon leaves implicit — thought as inner speech — its sister traditions state as definition. The Yogācāra scholastics define the pair thus: vitarkaḥ paryeṣako manojalpaḥ, vicāraḥ pratyavekṣako manojalpaḥ — applied thought is “searching mental-talk,” examination is “reviewing mental-talk.”[31] Sthiramati supplies the etymology — manaso jalpo manojalpaḥ … jalpo ‘rthakathanam, “mind’s talk: the stating of a meaning” — and, decisively for our question, illustrates the two functions with iti-labels: the searching mind proceeds kim etad iti (“what is this?”), the reviewing mind concludes idaṃ tad iti (“it is this”).[32] Yaśomitra calls vitarka the mind’s vāksamutthāpikā avasthā, its speech-prompting phase.[33] The honest qualifier: the Sarvāstivāda definition proper is cittaudārikatā, “grossness of mind,” without the talk — the explicit speech-equation is a Yogācāra formulation, though even the Kośa classes vitarka-vicāra as vāksaṃskāra, exactly as Dhammadinnā does.[34]
The same convergence holds for mindfulness itself. Sthiramati defines smṛti as abhilapanatā — the mind’s repeated re-articulation of its object, from abhi-√lap, a speech root; the Kośa’s satipaṭṭhāna discussion says the seen is retained yathādṛṣṭasya-abhilapanāt, “through re-speaking it as it was seen.”[35] The Pāli side has the twin: the Milindapañha gives sati the characteristic of apilāpana, illustrated by the king’s treasurer who recites the royal assets item by item — sati as the faculty that names and lists.[36] And where Pāli commentators coined sallakkheti for noting, the Sanskrit tradition turned the same root technical as upalakṣayati: the Śrāvakabhūmi has the practitioner note the components of experience upalakṣayaty upasthitayā smṛtyā, “with mindfulness present,” and the Kośa’s sixfold breath-method gives upalakṣaṇā, discernment-noting, a stage of its own.[37] The Śrāvakabhūmi also preserves the satipaṭṭhāna self-observation labels verbatim and first-person — asti me … saṃyojanam iti prajānāti — and governs the sixteen breath-resolves with śikṣate, precisely the Pāli sikkhati.[38]
The Similes
The images the tradition chose for the pair complete the picture. Nāgasena: vitakka has the characteristic of appanā — fixing, as a carpenter drives prepared wood into a joint — and vicāra of anumajjana, sustained stroking: “as when a bronze vessel is struck, the striking is vitakka, the reverberation is vicāra.”[39] The Visuddhimagga’s set is famous: the bell-strike and its resonance; the bird’s wingbeat and its glide; the bee’s dive toward the lotus and its circling above it; the hand that grips the tarnished dish and the hand that rubs it; the compass-pin that stands fixed and the pin that revolves.[40] The commentarial definition runs abhiniropana-lakkhaṇo vitakko — vitakka’s characteristic is the planting of the mind upon its object.[41] Strike, plant, fix, drive: every image renders vitakka a punctual act of placement, and vicāra the continuation that placement opens. A meditator who has used a label knows exactly what is being described: the note strikes; the observing resonates. The canonical images even supply the negative case: the thoroughbred in the stable, unlike the nag, does not brood ‘yavasaṃ yavasan’ti — “fodder, fodder!” — the very syntax of the repeating label, deployed as the picture of the wrong thing to rehearse.[42]
Part III: The Case Against, Weighed
An argument worth holding must survive its strongest opposition, and the census was built to find it. Four objections have real weight.
First, grammar, already conceded: iti is the only complementizer, so every instance is locally ambiguous, and the case rests wholly on distribution. This is a genuine methodological demotion — from proof-texts to statistics — and it should change how the thesis is argued forever. The 2008 post’s minimal pair (“had he meant as impermanent, he would have said aniccato“) also needs its qualification: the ablative construction does occur with the same verbs, 127 times in the canon against 75 for the quotative with the same predicates — though the 127 are heavily concentrated in one litany family (aniccato dukkhato rogato gaṇḍato…), so the alternation qualifies the pair without dissolving it.[43]
Second, the same frame carries content no one could note: the impossibility-knowledges of sutta length, future-tense contents — ‘kāyassa bhedā … sītībhavissantī’ti pajānāti, “he knows: ‘with the body’s breakup … all that is felt will grow cold’.”[44] The frame does double duty, and the practice-use must always be identified by shape — brevity, fixity, person, context — never by the frame alone.
Third, the ancient translators. The Chinese Āgamas render the formula overwhelmingly as plain knowing: the Madhyama-āgama’s signature is 行則知行 — “walking, then he knows walking” — and 入息長即知入息長, “breathing in long, he knows breathing in long,” dissolving the Indic first person into an object.[45] The control is sharp because these same translators possessed explicit quotative devices — 作是念 “he forms this thought,” over six hundred times in one collection; 念言 “he thinks, saying” — and declined to deploy them here.[46] Under the vocalization reading of the thesis this would be close to decisive. Under the vitakka reading its force narrows considerably: if the label is a thought with content, “knows X” is a defensible rendering, and what the translators actually flattened is the person — a loss the Ekottarika-āgama, alone, resists: 觀知我今息長, “he observes and knows: ‘my breath is now long'”; 自覺知我得樂痛, “he himself knows: ‘I feel a pleasant feeling’.”[47] One Āgama heard the first person and kept it.
Fourth, the commentators themselves demote the bare label. The aṭṭhakathā on the postures asks what ‘gacchāmī’ti pajānāti can mean, given that “even dogs and jackals, when walking, know that they are walking” — kāmaṃ soṇasiṅgālādayopi gacchantā ‘gacchāmā’ti jānanti — and answers that such knowing “does not abandon the perception of a being, does not remove the conceit of self, is neither a meditation subject nor satipaṭṭhāna-development.”[48] The sutta’s pajānāti, for the commentator, is the analysis that no walker exists apart from the process. Yet the same passage concedes the thesis its foundation in passing: ‘gacchāmī’ti cittaṃ uppajjati — “the thought ‘I am going’ arises, it produces the air-element…” — the quote is the content of a mental event; the commentator’s quarrel is only with anyone who would stop there.[49] The label, he insists, is the raw material of penetration, not its substitute. As a warning this stands against every mechanical use of noting; as an objection to the thesis in its vitakka form it has no force at all, for that form claims exactly what the commentator grants — the label arises as a thought, and the practice is what knowing does with it.
What would refute the thesis outright? If the label-lexicon were open-ended and clause-like, it would fall — it is not. If label-formulas ran on inside second-jhāna-and-above descriptions, the vitakka mechanism would fail — they do not. If the traditions defined vitakka in ways unconnected to speech, the mechanism would be a modern imposition — they define it as mental talk. The construction passes the tests that could have killed it.
Part IV: Stratigraphy — From Formula to Method
The census’s verb-table tells a history. Pajānāti owns the canonical formula — 1,666 canonical tokens — and then goes quiet: the commentaries do not compose new ‘X’ti pajānāti sentences; they inherit and gloss the old ones. Sallakkheti, “to mark, to note,” runs the other way: eight canonical tokens, every one of them in the Visuddhimagga, none in the four Nikāyas — against a hundred and fifteen in the aṭṭhakathā layer, where sallakkhetvā by itself occurs nearly six hundred times.[50] The subcommentary states the mature doctrine without embarrassment: yo pana aniccaṃ dukkhamanattāti tīṇi lakkhaṇāni sallakkheti — “one who notes the three characteristics” under their quoted names.[51] The Visuddhimagga meanwhile treats the samatha label as a name to be uttered: choosing among the designations of earth, yamicchati … taṃ vattabbaṃ — “whichever he wishes, that should be said” — and then ‘pathavī pathavī’ti developed, spoken and then, as the subcommentaries add, manasā eva punappunaṃ, “with the mind alone, again and again.”[52] It is a small irony of translation history that this kasiṇa instruction is the one place where translators have always kept the label literal — “earth, earth” — because there the practice was too obvious to dissolve.
The pipeline, then, is documented at every joint: canonical formulas whose shape is the shape of noting; a commentarial stratum that attaches to those very formulas a technical vocabulary of marking (sallakkhaṇā, vavatthāna) and an explicit pedagogy of labels chosen, spoken, interiorized, and retired; and a Burmese revival that took its verb — Mahasi’s sallakkheti — precisely from that stratum, and its exercises from those slots. Whether the Buddha “taught noting” in the strong sense no census can establish. That noting is the tradition’s own reading of its founder’s formulas, worked out by its own scholastics on both sides of the Indian ocean of transmission — sallakkhaṇā here, upalakṣaṇā and manojalpa there — the corpus now attests in detail. The Mahasi method is not an invention wanting for pedigree. It is the current form of the longest-running interpretation the formula has.
Part V: What This Means for the Meditator
For Buddhist studies, the census offers a construction: countable, classifiable, and consequential. For the practitioner, it offers something more personal — the discovery that the strangest-looking feature of one’s technique is the oldest.
First: the noting meditator practices inside the canon, not beside it. One who sits noting “rising,” “falling,” “hearing,” “wandering” is filling the very slots the texts fill — ‘gacchāmī’ti, ‘sarāgaṃ cittan’ti, ‘atthi me ajjhattaṃ kāmacchando’ti — with the very kind of content — brief, present-tense, first-person — the texts prescribe. The frequent charge that noting interposes “thinking” between the meditator and the moment has the psychology backwards, on the canon’s own account: the note is not added thinking but harnessed thinking, vitakka given one word of rope. Dhammadinnā’s doctrine dignifies the technique: since thought is inner speech anyway, the label merely takes the speech the mind was going to make and makes it serve attention.
Second: the label is an instrument, and the tradition’s own warning against fetishizing it should be posted in every meditation hall. Dogs and jackals know they are walking. The note that merely tags experience, without the seeing-through it exists to open — who walks? what is this “going” made of? — is the animal’s knowing with extra steps. The commentary and the Mahasi teachers say the same thing here, as heirs usually do: the note is for the sake of ñāṇamattāya, a mere measure of knowing, and the refrain’s -matta is a leash on the label as much as a license for it.
Third: the label is designed to die. Every stratum of the tradition — the vacīsaṅkhāra doctrine, the counting rule yāva vinā gaṇanāya, the thorn-simile, the Chinese ladder of silences — agrees that the labeling function extinguishes itself as concentration deepens, and that this is success, not failure. The practical corollary is worth stating plainly: a noting practice that cannot ever fall silent has stalled, and a concentration practice that despises the label has forgotten what carried it to the door. The gaṇanā doctrine’s counting-until-not-needed is perhaps the most humane piece of practice engineering the manuals contain, and it generalizes: note until noting is no longer needed; then let the silence do the noting.
Fourth: the formula is the Dhamma’s verifiability made concrete. When the Aṅguttara wants to show a skeptical brahmin that this teaching is sandiṭṭhika — visible for oneself, not a matter of faith or time — it does not offer him a doctrine to believe but a label to check: is there anger in you now? then you can know: ‘atthi me ajjhattaṃ…’ — “there is in me…”. A living Buddhism that wants to meet the modern demand for testability has, in this construction, its oldest asset: a canon whose central formulas are not creeds but observation protocols, written in the first person, in the present tense, one line long, waiting to be run. The translator can serve that Buddhism cheaply: keep the quotation marks. Let the English reader see what the Pāli reader has always seen — that at the heart of these texts, someone is being told what to know, and in which word to know it.
Conclusion
A census cannot recover intention; it can recover shape. The shape recovered here is of a canon saturated with short, fixed, first-person quotations lodged in the object-slot of verbs of knowing; a refrain that calls the bare label established mindfulness; a psychology that defines thought as unspoken speech and predicts — correctly, across three textual traditions — where the labels must fall silent; a scholasticism, Pāli and Sanskrit alike, that theorized the mechanism as mental talk and coined technical noting-verbs to name the practice; and a modern method whose vocabulary descends from that scholasticism by ordinary filiation. Against this stand the real concessions: grammar proves nothing locally, the frame carries other freight, and most ancient translators heard knowing rather than noting. The judgment this essay defends is that the concessions bound the claim without breaking it. The canonical formulas encode the drill; the manuals supply the theory of the drill; the noting traditions of the present practice both. What the editor of this blog saw in 2008 — an instruction staring out of a formula that translation had smoothed into description — was there to be seen, and is still there: seven thousand three hundred and eighty-four times.
A Note on Method and Data
All counts derive from a five-stage extraction and classification pipeline over the Chaṭṭha Saṅgāyana Tipiṭaka (Roman script, 217 XML files: mūla, aṭṭhakathā, ṭīkā, and para-canonical works), with parallel sweeps of the GRETIL Sanskrit corpus (784 TEI files) and CBETA editions of T 26, T 99, T 125, T 602, T 606, and T 607. Classification criteria are published as executable code; residual ambiguity (670 canonical tokens) is reported as a class, not resolved silently; a random spot-check of 36 classified tokens found one disputable assignment. The full token tables, distribution analyses, gloss surveys, Sanskrit and Chinese cross-check reports, and scripts are available as a data supplement. Citations below give CST file and paragraph, Taishō volume and line, or GRETIL file, as appropriate; every locus was machine-verified in the corpora.
Bibliography
Primary sources (digital editions used)
- Chaṭṭha Saṅgāyana Tipiṭaka, Roman script (Vipassana Research Institute XML). Canonical: Dīgha-, Majjhima-, Saṃyutta-, Aṅguttara-nikāya; Khuddaka (Suttanipāta, Paṭisambhidāmagga, Mahāniddesa, Milindapañha, Peṭakopadesa, Nettippakaraṇa); Vibhaṅga; Kathāvatthu. Commentarial: Sumaṅgalavilāsinī, Papañcasūdanī, Atthasālinī, and associated ṭīkās. Para-canonical: Visuddhimagga with Mahāṭīkā.
- GRETIL: Göttingen Register of Electronic Texts in Indian Languages, TEI corpus (2021 snapshot): Śrāvakabhūmi, Abhidharmakośabhāṣya with Vyākhyā, Abhidharmasamuccaya and Bhāṣya, Pañcaskandhaka with Sthiramati’s Vaibhāṣya, Triṃśikā-bhāṣya, Prajñāpāramitā corpus, Mūlasarvāstivāda-vinaya materials.
- CBETA XML P5: Madhyamāgama (T 26), Saṃyuktāgama (T 99), Ekottarikāgama (T 125), Da anban shouyi jing (T 602), Xiuxing daodi jing (T 606), Daodi jing (T 607).
Prior statements of the thesis
- Theravadin (ed.), “iti and sallakkheti,” theravadin.org, 28 March 2008.
- Theravadin (ed.), “Labels employed for noting in early Pitaka,” theravadin.org, 8 April 2009.
Data supplement
- The …’ti Census: token tables (7,384 rows), classification scripts, analysis tables, commentarial gloss survey, Sanskrit parallel tables and deep-dive report, Chinese cross-check report, label lexicon. (Repository accompanying this article.)
Notes
[1] Peṭakopadesa §23 (CST s0520m). The canonical Samādhi suttas deploy the same construction with yathābhūtaṃ intervening: “Samādhiṃ, bhikkhave, bhāvetha. Samāhito, bhikkhave, bhikkhu yathābhūtaṃ pajānāti. Kiñca yathābhūtaṃ pajānāti? ‘Cakkhu anicca’nti yathābhūtaṃ pajānāti…” (SN 35.99, CST s0304m §99); with the four truths as content at SN 56.1 (s0305m §1071). The Peṭakopadesa sentence is the formula’s tersest attested form and, as it happens, the sentence with which the 2008 essay opened.
[2] “iti and sallakkheti,” theravadin.org, 2008-03-28; the label-catalogue follows in “Labels employed for noting in early Pitaka,” 2009-04-08.
[3] Corpus: 217 CST Roman-script XML files. Extraction verbs and passes, classification rules, and all tables: data supplement, scripts/ and census-data/. Token total: 7,384; canonical (mūla) share: 2,702.
[4] On iti see PED s.v.: deictic particle marking off a statement “as the definite contents of (one’s own or other’s) thoughts,” “on the whole untranslatable (unless written as quotation marks).”
[5] DN 22 (CST s0102m §374) = MN 10 (s0201m §107); also MN 62, MN 118 (s0203m), SN 54 passim (s0305m), Vin pārājikā (vin01m).
[6] PED s.v. iti, quoted at length in the 2008 essay.
[7] Class table: data supplement, FINDINGS §3 and analysis-tables.md §1. Classes: L1 first-person label 843; L2-instr 367; L2-amb 670; E1 sati-refrain 20; E2 review 532; P-clause 63; P-long 192; unrecoverable 15.
[8] The exegetical idiom ‘…’ti daṭṭhabbaṃ / veditabbaṃ: 0 canonical vs 3,059 commentarial-and-later tokens.
[9] E.g. the ṭhānāṭṭhāna knowledges of MN 115, quotes of paragraph length under pajānāti; future-tense content at n. 44 below.
[10] 581 distinct normalized quotes fill the canonical label-slots; frequency table in the data supplement (analysis-tables.md §3); the families with full references in LABEL-LEXICON.md.
[11] First tetrad known: ‘dīghaṃ assasāmī’ti pajānāti (19 canonical tokens), ‘rassaṃ…’ (15); members 3–16 trained: ‘sabbakāyapaṭisaṃvedī assasissāmī’ti sikkhati etc.; 321 canonical sikkhati-tokens in all. Loci as n. 5.
[12] DN 22 (s0102m §375), MN 10 (s0201m §108), MN 119; with ‘ṭhitomhi’, ‘nisinnomhi’, ‘sayānomhi’.
[13] DN 22 (s0102m §380); Vibhaṅga Vedanānupassanāniddesa (abh02m §363); the sāmisa/nirāmisa pairs ibid.; the Dhātuvibhaṅga extension ‘kāyapariyantikaṃ vedanaṃ vedayāmī’ti pajānāti at MN 140 (s0203m §§363–364).
[14] Cittānupassanā: DN 22 (s0102m §381), MN 10, Vibhaṅga Cittānupassanāniddesa (abh02m; the Vibhaṅga inserts me: ‘sarāgaṃ me cittan’ti). Telepathy reuse: the sāmaññaphala sequence (s0101m) and Vism Cetopariyañāṇakathā (e0102n) apply the identical sixteen labels to another’s mind.
[15] DN 22 / MN 10, nīvaraṇa and bojjhaṅga sections; 97 canonical tokens of the atthi me / natthi me frame.
[16] Type-loci: SN 22.51 (s0303m); SN 35.99 (s0304m §99); the nibbānasappāya suttas (s0304m §§147–149); MN 28 extends the set to contact (s0201m).
[17] MN 22 (s0201m §241) and passim: 126 canonical tokens with samanupassati / daṭṭhabbaṃ / pajānāti, plus the wrong-view mirror ‘etaṃ mama…’ti samanupassati (37).
[18] MN 121 (s0203m §§177–180), with the graded ‘suññamidaṃ saññāgataṃ…’ti pajānāti ladder and the closing insight ‘yaṃ kho pana kiñci abhisaṅkhataṃ abhisañcetayitaṃ tadaniccaṃ nirodhadhamman’ti pajānāti (§183).
[19] AN 6.47–48 (s0403m2): sandiṭṭhika defined through the self-check label, second person.
[20] DN 22 / MN 10, closing refrain of each exercise; 20 canonical tokens across the four anupassanās (‘atthi kāyo’ti, ‘atthi vedanā’ti, ‘atthi cittan’ti, ‘atthi dhammā’ti).
[21] MN 44, Cūḷavedalla (s0201m §463 region).
[22] SN 41.6, Kāmabhū (s0304m).
[23] ‘Khīṇā jāti…’ passim; ‘atthi uttari nissaraṇan’ti pajānāti per attainment at MN 111, Anupada (s0203m §94).
[24] SN 48.39–40, Kaṭṭhopama (s0305m): on the arising of each faculty, ‘upekkhakosmī’ti / ‘sukhitosmī’ti / ‘dukkhitosmī’ti pajānāti.
[25] Third-jhāna pericope, passim: yaṃ taṃ ariyā ācikkhanti ‘upekkhako satimā sukhavihārī’ti.
[26] ‘Ananto ākāso’ti ākāsānañcāyatanaṃ upasampajja viharati; Vism arūpa parikamma ‘ākāso ākāso’ti … manasikātabbaṃ (e0101n, Paṭhamāruppa). 31 canonical label-tokens in formless-attainment contexts; discussion in FINDINGS §6.
[27] Vism VIII (e0101n §§223–224): ‘ekaṃ ekan’ti ādiṃ katvā yāva ‘dasa dasā’ti … gaṇetabbaṃ; the slow herdsman’s and the swift cowherd’s counting; terminus and purpose as quoted.
[28] SĀ, T 2, 99 at lb 0121b03.
[29] MĀ, T 1, 26 at lb 0561a08 (以覺觀為刺); AN 10.72, Kaṇṭaka (s0404m3 §72).
[30] Kv, Vacībhedakathā (abh03m3 §§326–327); attribution: Kv-a (abh03a): yesaṃ “sotāpattimaggakkhaṇe paṭhamaṃ jhānaṃ samāpannassa dukkhanti vācā bhijjatī”ti laddhi, seyyathāpi etarahi pubbaseliyādīnaṃ — “those whose doctrine is that at the moment of the stream-entry path, for one attained to the first jhāna, the utterance ‘dukkha’ breaks forth — as nowadays the Pubbaseliyas and others.”
[31] Abhidharmasamuccaya and Vasubandhu’s Pañcaskandhaka, s.vv. vitarka/vicāra (GRETIL files; verbatim passages in the data supplement’s Sanskrit deep-dive report).
[32] Sthiramati, Pañcaskandhaka-vaibhāṣā ad loc.
[33] Yaśomitra, Abhidharmakośa-vyākhyā ad AK ii.
[34] AKBh ad ii.33 region: vitarka as cittaudārikatā; vitarka-vicāra classed under vāksaṃskāra in the saṃskāra discussion — the Kośa’s echo of M i 301.
[35] Sthiramati, Triṃśikā-bhāṣya s.v. smṛti: abhilapanatā; AKBh smṛtyupasthāna discussion: yathādṛṣṭasyābhilapanāt, with Yaśomitra: abhilapyate smṛtyā udgṛhyate.
[36] Mil 2.1.13 (quoted with the treasurer simile in the Atthasālinī, abh01a): apilāpanalakkhaṇā sati; the treasurer sallakkhāpeti sāreti — “makes [the king] note, reminds him.”
[37] Śrāvakabhūmi: upalakṣayaty upasthitayā smṛtyā; AKBh vi, ānāpāna in six stages with upalakṣaṇā.
[38] Śrāvakabhūmi, satipaṭṭhāna section: asti me … saṃyojanam iti prajānāti etc.; the sixteen aspects with śikṣate (ti śikṣate, 43 occurrences in the Śbh files).
[39] Mil (s0518m §§13–14): vitakka appanālakkhaṇa with the carpenter; vicāra anumajjanalakkhaṇa with the struck bronze vessel (ākoṭanā / anuravanā).
[40] Vism IV (e0101n §71): the full set as listed.
[41] Atthasālinī (abh01a) and the fixed formula across some fifteen commentarial files: abhiniropanalakkhaṇo vitakko.
[42] AN 11.9, Sandha (s0404m4): na ‘yavasaṃ yavasan’ti jhāyati.
[43] Counts: 127 ablative-construction tokens (12 files) vs 75 quotative tokens for the anicca-family predicates with the same verbs; e.g. SN 35.53 cakkhuṃ aniccato jānato passato avijjā pahīyati (s0304m) beside SN 35.99’s quotative (n. 1).
[44] SN 12.51 and parallels (s0302m, s0303m): 7 canonical tokens.
[45] MĀ, T 1, 26: postures at lb 0582b20; breath at lb 0582c14; feelings at lb 0583c24. SĀ (T 2, 99): 覺知 and 善學 (for sikkhati) at lb 0206a28.
[46] 作是念: 630 occurrences in T 26; 念言: 25 in T 99 — freely used for narrative deliberation, never for the sensation-monitoring formulas. Full survey in the data supplement’s Chinese cross-check report.
[47] EĀ, T 2, 125: breath at lb 0556b01; feelings at lb 0568b28. The same collection flattens elsewhere (lb 0582a15), so the preservation is a trace, not a policy.
[48] Sv on DN 22 (s0102a §375) = Ps on MN 10 (s0201a §108).
[49] Ibid.: “‘gacchāmī’ti cittaṃ uppajjati, taṃ vāyaṃ janeti, vāyo viññattiṃ janeti…”
[50] Verb table: data supplement, FINDINGS §5. Sallakkheti quotative tokens: canon 8 (all Visuddhimagga: asubha-nimitta marking, ‘kālakato’ti, ‘mayhaṃ goṇā’ti), aṭṭhakathā 115; sallakkhetvā 594 across the commentarial corpus; vavatthapeti/vavatthapetvā 274.
[51] Mūlaṭīkā on the Dhammasaṅgaṇī commentary (abh01t).
[52] Vism IV (e0101n §§55–57): pathavīnāmesu yamicchati, yadassa saññānukūlaṃ hoti, taṃ vattabbaṃ. … pākaṭavaseneva pathavī pathavīti bhāvetabbaṃ; the mental-only continuation in the ṭīkā verse tradition (abh06t): karoto manasā eva punappunañca yogino.
