This one began with a proofreading session. I have been preparing a book for publication — more on that at the end, because the book has an unusual author — and its opening page prints a sentence this blog first discussed in 2008:
Samādhiṃ, bhikkhave, bhāvetha; samāhito, bhikkhave, bhikkhu rūpaṃ aniccan’ti pajānāti.
Develop concentration, monks. A concentrated monk knows form: “impermanent.”
I have read that sentence hundreds of times. This time a different word stopped me — not the quoted aniccan’ti, which the book is about, but the verb that governs it: pajānāti.
Strip the prefix and you get jānāti, the plain Pāli verb “to know.” Pāli is full of such pairs: a bare verb, and the same verb wearing a small prefix. Translators treat most of these prefixes as decoration, and pajānāti has accordingly been rendered “knows” or “understands” for a century and a half, as if the pa- were not there. But what if it is there for a reason? What if the pa- is a qualifier — the way German turns kennen into erkennen and greifen into begreifen — a little particle, obvious to the original listeners, that names a specific way of knowing?
The suspicion can be tested. Pāli has commentators who wrote down what they heard in words; it has a sister literature in Sanskrit; and it is now, thanks to work described below, a counted corpus, in which one can ask exactly where jānāti lives and where pajānāti lives. I spent the morning running the tests. Here is what came back.
Two knowings in English
Start with the distinction itself, because English has it too — it just uses separate words. There is the knowing of the courtroom: “he knows you cheated.” A fact is possessed; nothing about the knower’s state is claimed. And there is the knowing of the workshop or the clinic: the wine-taster discerns; the radiologist reads the scan the untrained eye merely sees. Same grammar, different faculty — a knowing that has a manner, that took training, that can be done well or poorly.
The question about pajānāti is whether Pāli splits this pair morphologically: bare jānāti for the courtroom, prefixed pajānāti for the trained discernment. If so, every satipaṭṭhāna formula has been flattened twice in translation — once when the quoted label was dissolved into a that-clause, and once when the verb of trained discernment was rendered as ordinary knowing.
The tradition answered first
The first test is to ask whether anyone in the tradition heard the prefix as meaningful — and it turns out Buddhaghosa answered the question directly, in the passage of the Visuddhimagga that defines wisdom itself:
Kenaṭṭhena paññā? Pajānanaṭṭhena paññā. Kimidaṃ pajānanaṃ nāma? Sañjānana-vijānana-ākāra-visiṭṭhaṃ nānappakārato jānanaṃ. (Vism XIV, CST e0102n §423)
In what sense is it paññā? In the sense of pa-jānana. And what is this pajānana? A knowing distinguished in mode from perceiving (sañjānana) and cognizing (vijānana) — a knowing in manifold ways.
Asked to define paññā, he reaches for its verb, and reads the prefix as the whole point: pajānana is not more knowing but differently-moded knowing — he glosses the pa- through pakāra, “mode, way.” Perception, consciousness, and wisdom all know; what separates them is how.
His illustration is aimed straight at the meditator. How do the three knowings differ? By which labels they can carry. Perception, he says, knows the object as “blue, yellow” — ‘nīlaṃ pītakan’ti — but cannot bring the labels ‘aniccaṃ dukkhamanattā’ti, “impermanent, suffering, not-self,” to penetration of the marks. Consciousness carries both the color-labels and the mark-labels, but cannot bring forth the path. Paññā carries all three. Buddhaghosa grades the mind’s knowings by the quotations they can sustain — which is to say, the fifth century already understood the two halves of this morning’s question as one: the quoted label and the prefixed verb belong to the same account of trained knowing.
The ears of the commentators
The second test: do the exegetes anywhere unpack a pa- compound of this verb in so many words? They do, on the closely related term every meditator knows — sampajañña, “clear comprehension,” which is saṃ + pa + jānana. Two commentaries, independently, gloss it:
Sampajaññaṃ: samantato pakārehi jānātī’ti attho. (Atthasālinī, abh01a; likewise the Niddesa commentary, s0515a)
Clear comprehension: the meaning is, “he knows from all sides, by modes.”
The saṃ- is heard as “from all sides,” and the pa- as pakārehi, “by modes, by ways.” Whatever a modern grammarian makes of that etymology, it settles the historical question: to the classical commentators, these prefixes were audible and semantic. They did not hear decoration.
The same commentarial tradition supplies a live contrast. Explaining “going, he knows: ‘going,'” the old commentary on the Satipaṭṭhāna Sutta raises its famous objection: even dogs and jackals, walking, know they are walking — and the verb it gives the animals is bare: ‘gacchāmā’ti jānanti. The practitioner’s knowing, it goes on, is something else: sampajānana, the mode-knowing, which sees the walking without a walker. Animal cognition gets the bare verb; the trained faculty gets the prefixed one. The distinction was not invented this morning; it was operating in fifth-century Ceylon.
Where each verb lives
The third test is the new one — the kind that only became possible this year. In preparing the book mentioned above, every quotative occurrence of these verbs in the Chaṭṭha Saṅgāyana canon was extracted and classified: every sentence of the shape ‘…’ti jānāti or ‘…’ti pajānāti, sorted by what fills the quote and where the sentence stands. If the prefix were empty, the two verbs should scatter alike. They do not scatter alike.
In the canonical layer, quotative pajānāti occurs 1,666 times, quotative jānāti 204 times. And the two live in different districts:
- In the Vinaya — the district of legal knowing, “he knows: ‘this monk is not suspended'” — bare jānāti outnumbers pajānāti 41 to 6.
- In the Saṃyutta Nikāya — the heartland of the practice formulas — the ratio reverses to 3 against 449. In the Dīgha, 4 against 172.
- The retrospective attainment-knowledges — “birth is destroyed,” “there is an escape beyond this” — run 421 to zero: not a single one is carried by the bare verb.
Complementary distribution is the linguist’s standard evidence that a difference in form marks a difference in function. Worldly and procedural knowing takes jānāti; the meditative registers take pajānāti, and the most exalted register takes it exclusively. The courtroom and the cushion each have their verb.
The Sanskrit career of the prefix
Fourth test: the sister tradition. In Sanskrit the same morphology gives pra-jñā — and there the prefixed form became the name of liberating knowledge, the prajñā of the prajñāpāramitā, while bare jñā derivatives stayed general. The Yogācāra meditation manuals preserve the satipaṭṭhāna formulas with the prefixed verb intact — asti me adhyātmaṃ … iti prajānāti, “he discerns: ‘there is in me…'” — in the same practice slots the Pāli gives to pajānāti. Two languages, one instinct about where the prefix belongs.
The case against
An argument this convenient needs its opposition stated at full strength. There are four objections, and two of them have real teeth.
Prefix bleaching. Languages wear prefixes smooth. English “eat up” adds little to “eat”; plenty of Pāli verbal prefixes are, by common scholarly judgment, semantically light. The general phenomenon is real — which is why the argument cannot rest on morphology alone, and rests instead on the convergence of native testimony, exegetical practice, and distribution. A bleached prefix does not get glossed pakārehi by two commentators, and it does not sort itself 421-to-zero across a register.
The exceptions exist. Bare jānāti does occasionally carry practice-labels. The clearest case is the Anumāna Sutta’s self-review: “if, reviewing, a monk knows (jānāti): ‘I am angry, overcome by anger,’ he should strive…” (MN 15, s0201m). A hard version of the thesis — pajānāti always and only for practice — is false, and should not be defended. The defensible claim is distributional: the canon strongly prefers the prefixed verb in practice registers and reserves the review-knowledges for it absolutely. Living languages implement meaning as strong preference more often than as exceptionless law.
Formulaic inheritance. Perhaps the practice formulas simply congealed early around pajānāti, and the distribution reflects frozen phrasing, not live choice. This deserves respect — much of the canon is formula. But it explains less than it seems to: the question would simply move one step back, to why the formula-makers reached for the prefixed verb when building the practice register in the first place. Frozen phrases preserve old choices; they do not erase them.
Meter and cadence. A four-syllable verb sits differently in a clause than a three-syllable one, and prose rhythm has preferences. Possibly a contributing force; no one has shown it sufficient, and it is silent about the Vinaya’s opposite preference.
Weighed together: the objections trim the thesis to its defensible size and leave its core standing. The prefix was audible, the commentators said what it meant, and the corpus distributes the two verbs the way a semantic difference distributes.
What changes for the meditator
Suppose the trimmed thesis is right. What follows, on the cushion?
Start with a distinction the texts themselves draw, because there are three different ways a label can ride experience, and only one of them is the instruction.
A label can ride as a caption: the word merely registers what perception has already delivered. “Hearing, hearing” as bare tagging — the moment filed, nothing seen. This is the jackal’s knowing, on the commentary’s own telling: even animals, walking, know they are walking. The sutta declined to use their verb.
A label can ride as a mantra: the word repeated for its own momentum, with no one looking at anything. The canon knows this failure mode and mocks it — the nag at the feeding-trough broods “fodder, fodder!” — and the Visuddhimagga’s counting rule exists to prevent it: count only until mindfulness stays without counting. (Repetition itself is not the failure: the samatha wing prescribes repetition with an object — “earth, earth,” the counting numbers, and the later tradition’s buddho, buddho stand in that lineage, one word planting the mind again and again on a single object until it unifies, the word thinning as absorption nears. The failure is the word running with no object at all — repetition that never converges, toward either stillness or seeing.)
And a label can ride as a probe: the word deployed at the object, as the vehicle of a discernment that goes in. This is what the grammar of the formula orders. Its structure is three-part: the occurring (breathing in long), the word (“long”), the mode (pajānāti — the trained knowing). The tradition’s similes describe the mechanics precisely. Vitakka’s mark is abhiniropana — the label plants the mind on its object. The compass has a fixed pin and a revolving pin. And in the simile of the tarnished dish, one hand grips while the other rubs: the label is the gripping hand. It polishes nothing itself; it holds the thing still so discernment can work. Not a caption, not a mantra — an anchor for seeing deeply. German has the right verb for what happens next: begreifen, to grasp-into-understanding.
This raises the practical question a meditator will already have formed: is the instruction “attach this label when you notice the thing”? Or is it “train yourself to see the thing under this description, until the seeing itself changes” — a re-programming? The canon’s answer is both, on different shelves, and the tense of the formula tells you which shelf you are on. The pajānāti formulas are occasioned acts, present tense, triggered by the event: when the mind has lust, he discerns: “a mind with lust.” The deliberate re-seeing lives in the other frames: ‘aniccānupassī assasissāmī’ti sikkhati — “he trains: ‘contemplating impermanence I will breathe in'” — a way of seeing installed in advance, by rehearsal, in the future tense. And the Saṃyutta states the payoff of the standing version in one line: aniccato jānato passato avijjā pahīyati — “for one knowing and seeing as impermanent, ignorance is abandoned” (SN 35.53). The momentary label is not itself the reprogramming. Ten thousand occasioned acts of discernment are how the reprogramming happens — and where the tradition wanted the new default pre-loaded rather than slowly accumulated, it wrote future-tense scripts and said sikkhitabbaṃ, “thus should you train.” One instrument; the tenses are its settings.
Buddhaghosa’s grading then gives the practitioner a live diagnostic. Perception carries “blue, yellow”; paññā carries “impermanent” through to penetration. Anyone can watch the difference in their own noting: the same word can ride as caption — filed, inert — or as probe, catching the arising edge, the passing edge, the not-mine of the sound. Same quote; different verb doing the knowing. The formulas ask for the second, and the difference is trainable, which is presumably why the canon wrapped its instructions in a verb that names the trained mode.
Last, a friendlier view of a hard old word. Sampajañña, “clear comprehension,” has always been the satipaṭṭhāna refrain’s mystery ingredient. Unpacked as the commentators unpack it — knowing from all sides, by modes — it stops being a fourth thing to do and becomes the name of doing the first thing well: the label carried by discernment rather than recognition, from as many sides as the moment offers.
The census figures above come from the corpus study behind a book now in final preparation — Hidden in Plain Sight: The Buddha’s Meditation Instructions You Were Never Shown — which restores the quotation marks this blog began arguing for in 2008, catalogues the canon’s sixty labels, its standing reviews, and its 238 prescribed interior monologues, and follows that library from the suttas through the commentators to the Burmese noting tradition. The pajānāti finding arrived while proofreading its page five, too late for the first printing; it will open the second.
The particle was the message all along. It just turned out there were two of them.
— Quentin I. Marks
References
All Pāli cited from the Chaṭṭha Saṅgāyana Tipiṭaka (Vipassana Research Institute, Roman-script edition); CST file and paragraph given where applicable.
- Peṭakopadesa §23 (CST s0520m): samādhiṃ bhikkhave bhāvetha; samāhito bhikkhave bhikkhu rūpaṃ aniccan’ti pajānāti. Canonical parallels with yathābhūtaṃ: SN 35.99 (s0304m §99); SN 56.1 (s0305m §1071).
- Visuddhimagga XIV (CST e0102n §423): kenaṭṭhena paññā? pajānanaṭṭhena paññā… sañjānana-vijānana-ākāra-visiṭṭhaṃ nānappakārato jānanaṃ, with the grading of saññā, viññāṇa, and paññā by the labels each can carry.
- Atthasālinī (abh01a) and Saddhammapajjotikā (s0515a), s.v. sampajañña: samantato pakārehi jānātī’ti attho.
- Sumaṅgalavilāsinī on DN 22 (s0102a §375) = Papañcasūdanī on MN 10 (s0201a §108): kāmaṃ soṇasiṅgālādayopi gacchantā ‘gacchāmā’ti jānanti… — the animals’ bare verb; the practitioner’s knowing routed to sampajānana.
- Distribution counts: quotative pajānāti 1,666 vs jānāti 204 in the canonical (mūla) layer; Vinaya 41:6 for the bare verb; SN 3:449 and DN 4:172 for the prefixed; attainment-reviews 421:0. Method and dataset as documented in the book’s Appendix D (census of 7,384 quotative tokens, classification rules published as code).
- MN 15, Anumānasuttaṃ (s0201m): sace… paccavekkhamāno evaṃ jānāti ‘kodhano khomhi kodhābhibhūto’ti… — the bare-verb counter-example.
- SN 35.53 (s0304m): cakkhuṃ aniccato jānato passato avijjā pahīyati.
- AN 11.9, Sandhasuttaṃ (s0404m4): na ‘yavasaṃ yavasan’ti jhāyati.
- Visuddhimagga VIII (e0101n §§223–224): the counting labels and their terminus, yāva vinā gaṇanāya assāsapassāsārammaṇe sati santiṭṭhati.
- Śrāvakabhūmi (GRETIL): asti me adhyātmaṃ… iti prajānāti — the prefixed verb in the Sanskrit satipaṭṭhāna formulas.
- Aṅguttara ṭhāna: MN 21 similes and further practice files as collected in the book’s Appendices A–B.
