The lexicon — lexicon

RENDERED FROM STRUCTURED DATA — this page is not a markdown file. Source: lexicon/terms/method-evidence.json in uni-cookbook at 575fc93d9d31. The sha256 recorded for this page is the digest of that JSON, not of this text. The renderer is generators/derive_docs.cjs; it copies fields and adds no facts. Every value below is a field of the source, verbatim, including the ones that read NOT RUN and PENDING.

The five-register renderings (Sanskrit / Latin / English / Hindi / Spanish) for the concepts CONCEPTS.json routes to the method domain. Each entry translates the concept's DEFINITION, never its English headword; the English headword is a label. NOTHING HERE IS CLAIMED PERFECT. Per rule 18.14 of the operator's charter ('Do not claim perfect translation'), no rendering is admitted as correct by being written here: it is admitted only against a cited dictionary headword plus a PASSING back-translation test, and no back-translation test in this file has been run. Every entry therefore carries reviewer = 'NONE - PENDING EXPERT REVIEW'. This file raises no UNI rung, discharges no gate, and states no fact about nature.

authored 2026-07-15 · counts: {"concepts_in_file":21,"all_store_TRUE":true,"back_translation_tests_run":0,"expert_reviews_completed":0,"status_tally_note":"Tally is deliberately not asserted here as a summary number; each register carries its own status and a reviewer must count them rather than trust a headline."}

Falsifiers for this file

• A concept_id here is absent from CONCEPTS.json, or its definition is not verbatim, or a concept whose domain == 'method' is missing. Expected membership: exactly the 21 ids listed in this file.
• Any register marked CITED whose authority does not carry that headword, or does not carry it in the quoted sense. That entry is a pseudopropheta and must be struck to PENDING-CITATION.
• Any register marked COINED lacking a morphological derivation.
• A back_translation_test is recorded as PASSING anywhere in this file. None has been run; any such record is a defect.
• An ambiguity_note or forbidden_mistranslation asserts a metaphysics as settled rather than recording that the register's word imports one.
• A rendering merges the TRUE and HONEST stores - e.g. any register for CALIBRATION that would license calibrating testimony.

FALSIFIER — falsifier

Store: TRUE · Domain: method · Confidence: LOW for Latin/Hindi/Spanish (all unattested constructions; the counterfactual is unhandled). MEDIUM for Sanskrit bādhaka, whose attested 'annulling' sense is close but whose doctrinal freight is heavy. · Reviewer: NONE - PENDING EXPERT REVIEW

A stated observation that would refute a claim, written before the run, without which surprise cannot be computed and the claim is inadmissible.

register	form	literal gloss	status
sanskrit	बाधक (bādhaka)	setting aside, suspending, annulling; opposing, hindering	CITED
latin	observātiō refūtāns	the refuting observation - the observation that (as it happens) refutes	COINED
english	falsifier	A stated observation that would refute a claim, written before the run, without which surprise cannot be computed and the claim is inadmissible.	—
hindi	खंडक प्रेक्षण ()	refuting observation	PENDING-CITATION
spanish	observación falsadora	falsifying observation (the observation that would refute)	PENDING-CITATION

Forbidden mistranslations

register	form	why it is wrong
		L&S s.v. falsĭfĭcus: 'that acts falsely,' Plaut. Mil. 2, 2, 36. The word means acting deceitfully - a FORGER or a liar. A falsificator would name a person who fakes data: the precise opposite of the instrument that catches them.
		khaṇḍana is the ACT of refuting (and the title of Śrīharṣa's Advaita polemic), not the pre-registered observation that would do the refuting. It names the debater's move, losing 'written before the run'.
		An objection is raised AFTER seeing the result. The falsifier is written BEFORE it. Drop the temporal order and the concept collapses into the narration failure its own definition warns against.
		Means 'forger, counterfeiter' in ordinary Spanish. Same defect as Latin falsificator, and more strongly marked.

Back-translation test — status: NOT RUN: no reader has been asked

RECEIPT — receipt

Store: TRUE · Domain: method · Confidence: MEDIUM on the attestations (all four are solid quotations). LOW on fitness: all four are financial or forensic words and none carries the loop-closing sense. This entry is a strong candidate for rejection in favour of a coinage. · Reviewer: NONE - PENDING EXPERT REVIEW

The captured observation that closes a turn of the loop -- an audit id and UTC for live state, file:line for code, a run log for tests -- against which the prior prediction is compared.

register	form	literal gloss	status
sanskrit	लेख्य (lekhya)	a writing, document; (in law) any written document	CITED
latin	apocha	the receipt of a creditor acknowledging the payment of a debt	CITED
english	receipt	The captured observation that closes a turn of the loop -- an audit id and UTC for live state, file:line for code, a run log for tests -- against which the prior prediction is compared.	—
hindi	रसीद ()	receipt; acknowledgement of receipt	CITED
spanish	recibo	written, signed acknowledgement that something was received	CITED

Forbidden mistranslations

register	form	why it is wrong
		The Sceptics' ἐποχή/ἀποχή is the withholding of assent. UNI's receipt is where assent is finally EARNED against a prediction. Same letters, opposite epistemic act.
		McGregor p. 857 marks '~ करना' as colloq. 'to strike, to land (a blow: on को)'. The verb phrase means to hit someone.
		McGregor p. 662 gives प्रमाण-पत्र as 'certificate; diploma; credentials'. A certificate ASSERTS a status on someone's authority. A receipt RECORDS an observation. 'certified' is also banned in author voice.
		An invoice - a demand for payment, i.e. an obligation opened, not discharged. Exactly inverts the concept.

Back-translation test — status: NOT RUN: no reader has been asked

EVIDENCE_CLASS — evidence class

Store: TRUE · Domain: method · Confidence: LOW. pramāṇa is attested and close, but the doctrinal freight is severe and the ceiling property is carried by no register. The Spanish false friend is confirmed and disqualifying for the obvious rendering. · Reviewer: NONE - PENDING EXPERT REVIEW

The A-to-U label recording HOW a claim is known, above which no claim may ever be raised.

register	form	literal gloss	status
sanskrit	प्रमाण (pramāṇa)	a means of acquiring pramā (certain knowledge); any proof or testimony or evidence; right measure, standard, authority	CITED
latin	classis probātiōnis	the class/rank of proof	COINED
english	evidence class	The A-to-U label recording HOW a claim is known, above which no claim may ever be raised.	—
hindi	प्रमाण-कोटि ()	class/category of evidence	PENDING-CITATION
spanish	clase de prueba	class of proof/evidence	COINED

Forbidden mistranslations

register	form	why it is wrong
		The worst error available in this file. RAE s.v. evidencia: '1. f. Certeza clara y manifiesta de la que no se puede dudar.' Spanish evidencia primarily means CERTAINTY BEYOND DOUBT. A label whose entire job is to record how weakly something is known would be rendered as a label asserting indubitable certainty - the exact inversion. Use prueba.
		L&S s.v. classis: in military language 'the fleet'. Unqualified, classis reads as a navy.
		Bare pramāṇa asserts 'this is a valid means of knowledge / this is authoritative'. UNI's class asserts the opposite direction: it is a ceiling, not a warrant. It also silently commits UNI to one school's enumeration.
		The class records HOW a claim is known (the kind of access), not how GOOD it is. A Class-A runtime observation is not 'better quality' than a Class-E test pass; it is a different kind of access, and that is why the authority ordering exists.

Back-translation test — status: NOT RUN: no reader has been asked

CLASS_AUTHORITY_ORDERING — class authority ordering

Store: TRUE · Domain: method · Confidence: LOW across all registers. Nothing here is attested; the Latin and Hindi are honestly incomplete. This is the weakest entry in the file and should be treated as a placeholder. · Reviewer: NONE - PENDING EXPERT REVIEW

M9's rule that Class-B tool state overrides Class-G own narrative and Class-A runtime observation overrides Class-E test pass, so that a passing test does NOT satisfy a criterion demanding a runtime observation.

register	form	literal gloss	status
sanskrit	प्रमाण-बल-क्रम (pramāṇa-bala-krama)	the order of force among evidence classes	COINED
latin	ōrdō auctōritātis classium	the order of authority of the classes	PENDING-CITATION
english	class authority ordering	M9's rule that Class-B tool state overrides Class-G own narrative and Class-A runtime observation overrides Class-E test pass, so that a passing test does NOT satisfy a criterion demanding a runtime observation.	—
hindi	—	—	NOT-ATTEMPTED
spanish	orden de prelación de las clases de prueba	order of precedence among evidence classes	PENDING-CITATION

Forbidden mistranslations

register	form	why it is wrong
		Priority is about what to do FIRST. M9 is about what OVERRIDES what when they conflict. A test run first and a runtime observation run second does not change which wins.
		Makes the rule say 'whoever has more standing wins', reintroducing exactly the Class-G own-narrative authority that M9 exists to demote.
		Several schools rank scriptural testimony at or above perception. M9 ranks runtime observation ABOVE narrative. A rendering inside that apparatus can silently invert the rule.

Back-translation test — status: NOT RUN: no reader has been asked

LEDGER — ledger

Store: TRUE · Domain: method · Confidence: MEDIUM on attestation for Latin and Hindi (both solidly quoted). LOW on fitness for Latin, whose famous collocation inverts the append-only property. Sanskrit honestly not attempted. · Reviewer: NONE - PENDING EXPERT REVIEW

The append-only single source of truth whose four states are PASS, FAIL, NEGATIVE and PENDING, and which wins wherever prose disagrees with it.

register	form	literal gloss	status
sanskrit	—	—	NOT-ATTEMPTED
latin	tabulae	a book of account, account-books	CITED
english	ledger	The append-only single source of truth whose four states are PASS, FAIL, NEGATIVE and PENDING, and which wins wherever prose disagrees with it.	—
hindi	बही ()	account-book, ledger	CITED
spanish	libro mayor	ledger (the principal account-book)	PENDING-CITATION

Forbidden mistranslations

register	form	why it is wrong
		L&S: new account-books substituted for the old ones so that debts were cancelled. Names the abolition of the record. For an append-only ledger this is the single most damaging phrase in Latin.
		RAE s.v. registro gives 'Acción de registrar' and 'Lugar desde donde se puede registrar o ver algo' - the act of registering or a place to inspect from; it also strongly carries 'search' (a police registro is a search). It does not carry account-book.
		A log is a narrative record of events. The ledger holds ADJUDICATED STATES (PASS/FAIL/NEGATIVE/PENDING) and outranks prose. Rendering it as a diary makes it Class-G own narrative - exactly what M9 demotes.
		L&S: a board or plank; a writing-tablet. Only the PLURAL carries the account-book sense. The singular names the physical object.

Back-translation test — status: NOT RUN: no reader has been asked

CALIBRATION — calibration

Store: TRUE · Domain: method · Confidence: MEDIUM for Latin and Sanskrit as coinages: both members are cited and the fence word does real work, and śodhana's attested subtraction sense is a genuine support. LOW for Hindi/Spanish, honestly incomplete. The 'prior wording stays on record' clause is carried by NO register and is a known gap. · Reviewer: NONE - PENDING EXPERT REVIEW

Moving wording DOWN to the measured value -- never up -- and keeping the prior wording on record, so that the fence gets louder under pressure rather than wider.

register	form	literal gloss	status
sanskrit	अपकर्ष-शोधन (apakarṣa-śodhana)	downward correction	COINED
latin	correctiō deorsum	correction downwards	COINED
english	calibration	Moving wording DOWN to the measured value -- never up -- and keeping the prior wording on record, so that the fence gets louder under pressure rather than wider.	—
hindi	—	—	NOT-ATTEMPTED
spanish	corrección a la baja	downward correction (of a stated figure or wording)	PENDING-CITATION

Forbidden mistranslations

register	form	why it is wrong
		L&S s.v. correctio: the rhetorical figure substitutes 'a stronger or more significant' word. Unfenced, the Latin licenses correcting UPWARD, which is precisely what UNI's calibration prohibits.
		MW s.v. शोधन also gives 'excrement, ordure' and 'expiation', 'retaliation, punishment'. Bare śodhana can read as purgation, penance or punishment. It also means 'purifying', which imports the idea that the earlier wording was DIRTY rather than merely uncalibrated - and the prior wording must stay on record with dignity, not be cleansed away.
		Names adjusting an instrument against a standard, symmetric in direction. UNI's calibration adjusts WORDING and only ever downward. The instrument reading also invites the catastrophic idea that a person's lived report could be 'calibrated'.
		HONEST signals are never calibrated. This crosses the TRUE/HONEST sovereign split - the cardinal sin. Not a nuance: a rendering that permits this object is disqualified outright.

Back-translation test — status: NOT RUN: no reader has been asked

NEGATIVE — negative

Store: TRUE · Domain: method · Confidence: MEDIUM-HIGH on the Latin attestation and structurally the best match in the file; MEDIUM on its fitness given the electorate problem. Three registers honestly not attempted rather than padded. · Reviewer: NONE - PENDING EXPERT REVIEW

A held UNI run that missed its pre-registered bar: a result UNI earned, published front-and-center as the credibility, and never retried until it wins.

register	form	literal gloss	status
sanskrit	—	—	NOT-ATTEMPTED
latin	repulsa	a refusal, denial, repulse in soliciting for an office; a defeat in a candidacy	CITED
english	negative	A held UNI run that missed its pre-registered bar: a result UNI earned, published front-and-center as the credibility, and never retried until it wins.	—
hindi	—	—	NOT-ATTEMPTED
spanish	—	—	NOT-ATTEMPTED

Forbidden mistranslations

register	form	why it is wrong
		A failure is something that went wrong. A NEGATIVE is a run that went RIGHT and returned an unwelcome answer. The definition calls it a result UNI earned and publishes it as the credibility.
		Reads as 'the test found nothing / you don't have it'. Loses the pre-registered bar entirely, and a negative test finding is not a missed bar.
		Breaks the two-sovereign-vocabularies fence. Nature's class for that is SUPERSEDED. One token may not serve both, and NATURE-LEDGER.md governs the other.
		MW: the position of being unfit to carry on an argument from impossibility of agreeing about first principles. It is a debate forfeit over premises, not a measured miss of a stated bar. Rejected on the evidence, not on taste.

Back-translation test — status: NOT RUN: no reader has been asked

BASELINE — tuned baseline

Store: TRUE · Domain: method · Confidence: MEDIUM for the Latin coinage - aemulus carries the effort-symmetry unusually well and both members are cited - with the bad-sense trap noted. Three registers honestly not attempted. · Reviewer: NONE - PENDING EXPERT REVIEW

A strong comparison tuned with the same effort spent on the method itself, since an untuned baseline measures the author's enthusiasm rather than the mechanism.

register	form	literal gloss	status
sanskrit	—	—	NOT-ATTEMPTED
latin	norma aemula	the rival standard - a standard that strives to equal	COINED
english	tuned baseline	A strong comparison tuned with the same effort spent on the method itself, since an untuned baseline measures the author's enthusiasm rather than the mechanism.	—
hindi	—	—	NOT-ATTEMPTED
spanish	—	—	NOT-ATTEMPTED

Forbidden mistranslations

register	form	why it is wrong
		Names the pre-intervention measurement you improve away from. M7's baseline is a competing METHOD tuned to beat you, not a starting level. Different object.
		L&S records the hostile sense and the ecclesiastical metonymy 'an enemy'. A baseline built as an enemy gets built weak - the exact failure M7 names, where the comparison measures the author's enthusiasm.
		A control holds the intervention out. A tuned baseline is a fully-tuned ALTERNATIVE METHOD competing on the same metric. Conflating them also collides with POSITIVE_CONTROL, a separate concept in this file.
		'Tuned with the same effort' IS the concept. An untuned reference is exactly the defective object the definition warns against.

Back-translation test — status: NOT RUN: no reader has been asked

DISCRIMINATOR — load-bearing discriminator

Store: TRUE · Domain: method · Confidence: LOW-MEDIUM. vyāvṛtti is attested and reasonably close but nominalises the action and carries apoha freight; the Latin coinage is defensible; 'load-bearing' is unhandled everywhere. · Reviewer: NONE - PENDING EXPERT REVIEW

A manipulation -- shuffling labels, swapping markers, ablating the mechanism to zero -- that must COLLAPSE the claimed gain, since a gain surviving the destruction of its own mechanism was never the mechanism's.

register	form	literal gloss	status
sanskrit	व्यावृत्ति (vyāvṛtti)	exclusion, rejection, removal; discrimination, distinction	CITED
latin	experīmentum discernēns	the distinguishing test - the trial that separates	COINED
english	load-bearing discriminator	A manipulation -- shuffling labels, swapping markers, ablating the mechanism to zero -- that must COLLAPSE the claimed gain, since a gain surviving the destruction of its own mechanism was never the mechanism's.	—
hindi	—	—	NOT-ATTEMPTED
spanish	—	—	NOT-ATTEMPTED

Forbidden mistranslations

register	form	why it is wrong
		MW: in Vedānta, the power of separating Spirit from world, truth from untruth, reality from illusion. Imports Advaita soteriology wholesale and makes a label-shuffling control read as liberating insight.
		L&S: 'that which separates or divides two things' - a difference that obtains, and also 'a decisive point, critical moment' and even 'the dividing membrane, the midriff'. Names a state, not a manipulation you perform.
		The dominant modern sense of discriminar in Spanish is to discriminate AGAINST (unjust treatment). Unfenced, the term reads as a social-justice word.
		Those ask whether the result SURVIVES perturbation, and survival is the good outcome. The discriminator demands the gain COLLAPSE, and survival is the bad outcome - it proves leakage. The rendering inverts the pass condition.

Back-translation test — status: NOT RUN: no reader has been asked

ABLATION — ablation

Store: TRUE · Domain: method · Confidence: MEDIUM-HIGH on attestation (lopa and ablatio are both solidly quoted). LOW on completeness: the purpose clause and the computed-residual fence are carried by no register, and Spanish is blocked by a confirmed and serious false friend. · Reviewer: NONE - PENDING EXPERT REVIEW

Removal of a mechanism in order to measure its contribution, whose number must be a computed residual and never a literal typed in by hand.

register	form	literal gloss	status
sanskrit	लोप (lopa)	(in grammar) dropping, elision; want, deficiency, absence, disappearance	CITED
latin	ablātiō	a taking away	CITED
english	ablation	Removal of a mechanism in order to measure its contribution, whose number must be a computed residual and never a literal typed in by hand.	—
hindi	—	—	NOT-ATTEMPTED
spanish	—	—	NOT-ATTEMPTED

Forbidden mistranslations

register	form	why it is wrong
		RAE s.v. ablación: '1. f. Med. Extirpación de cualquier parte del cuerpo. 2. f. por antonom. En algunas culturas, extirpación ritual del clítoris. 3. f. Der. Sacrificio o menoscabo de un derecho.' By antonomasia, ablación in Spanish means female genital mutilation. Using it unqualified for an ablation study is a severe register error, not a stylistic one.
		The grammatical ablative is a case name from the same root. A Latin reader may hear a declension, not an experiment.
		Surgical excision removes tissue to CURE. Ablation removes a mechanism to MEASURE it. The purpose clause is the concept; a curative reading deletes it.
		Outside grammar lopa means 'injury, destruction' and 'violation, transgression (of a vow or duty)'. The technical Pāṇinian sense is the only one that fits, and it must be explicitly fenced.

Back-translation test — status: NOT RUN: no reader has been asked

PRE_REGISTRATION — pre-registration

Store: TRUE · Domain: method · Confidence: MEDIUM-HIGH for Sanskrit and Hindi pratijñā (well attested, structurally apt, with a known moralising drift) and MEDIUM for the Latin coinage, whose census-registration sense is genuinely close. Spanish owed. · Reviewer: NONE - PENDING EXPERT REVIEW

Writing the bar, the margin, the threshold, the named ablation and the stopping rule BEFORE the build, because reasonable data-contingent analytic choices inflate false positives with no deliberate p-hacking at all.

register	form	literal gloss	status
sanskrit	प्रतिज्ञा (pratijñā)	(in logic) the proposition to be proved, the first member of the five-membered Nyāya syllogism; a declaration, promise, vow	CITED
latin	professiō praevia	a prior public declaration entered on a register	COINED
english	pre-registration	Writing the bar, the margin, the threshold, the named ablation and the stopping rule BEFORE the build, because reasonable data-contingent analytic choices inflate false positives with no deliberate p-hacking at all.	—
hindi	प्रतिज्ञा ()	assertion, declaration; assurance; vow; promise	CITED
spanish	—	—	NOT-ATTEMPTED

Forbidden mistranslations

register	form	why it is wrong
		Moralises the concept. Pre-registration is not a pledge of good behaviour; the definition explicitly locates the harm in REASONABLE choices made with no deliberate p-hacking. A vow implies the danger is bad faith, which is the wrong diagnosis and licenses 'I am honest, so I need not pre-register'.
		L&S also gives 'A business or profession which one publicly avows' - 'professio bene dicendi,' Cic. de Or. 1, 6, 21. Unfenced it reads as one's occupation, and it loses the time order that praevius supplies.
		pratijñā's logical sense is the proposition to be proved. Pre-registration fixes the BAR, the MARGIN, the THRESHOLD, the named ABLATION and the STOPPING RULE - the decision procedure, not the claim. A rendering that names only the thesis loses the stopping rule, which is where the inflation actually enters.
		RAE s.v. registro: 'Acción de registrar'; it also strongly carries 'search'. Reads as administrative enrolment, and drops 'before the build' entirely.

Back-translation test — status: NOT RUN: no reader has been asked

HELD_SET — held set

Store: TRUE · Domain: method · Confidence: LOW. Only Latin attempted, as a coinage with one member uncited; the once-only sentinel - the actual discriminating property - is carried by nothing. Three registers honestly not attempted rather than invented. · Reviewer: NONE - PENDING EXPERT REVIEW

Data touched exactly ONCE behind a seal-before-scoring step and a once-only sentinel, since a held set scored twice is a training set with a good reputation.

register	form	literal gloss	status
sanskrit	—	—	NOT-ATTEMPTED
latin	pars obsignāta	the sealed portion	COINED
english	held set	Data touched exactly ONCE behind a seal-before-scoring step and a once-only sentinel, since a held set scored twice is a training set with a good reputation.	—
hindi	—	—	NOT-ATTEMPTED
spanish	—	—	NOT-ATTEMPTED

Forbidden mistranslations

register	form	why it is wrong
		The seal is not about concealment - the data may be perfectly well known. It is about foreclosing a second scoring. Secrecy is the wrong mechanism and would license 'nobody looked, so it is fine'.
		Reserving data says nothing about how many times it may be scored. The definition's whole force is 'touched exactly ONCE' - a set scored twice is a training set with a good reputation.
		'held-PASS' is banned in author voice by the honesty rail. The set being held says nothing about the outcome.

Back-translation test — status: NOT RUN: no reader has been asked

BOUND_NOT_POINT — the bound, not the point estimate

Store: TRUE · Domain: method · Confidence: LOW. The attestations are genuine but they attest ordinary boundary words; the concept is a verdict RULE and is not rendered by any of them. Recorded as a known structural gap, not as a near miss. · Reviewer: NONE - PENDING EXPERT REVIEW

M2's verdict rule: the verdict is the confidence-interval bound that EXCLUDES the threshold, never the point estimate.

register	form	literal gloss	status
sanskrit	अवधि (avadhi)	a term, limit; conclusion, termination	CITED
latin	terminus excludēns	the excluding boundary	COINED
english	the bound, not the point estimate	M2's verdict rule: the verdict is the confidence-interval bound that EXCLUDES the threshold, never the point estimate.	—
hindi	सीमा ()	limit, boundary; frontier	CITED
spanish	—	—	NOT-ATTEMPTED

Forbidden mistranslations

register	form	why it is wrong
		The verdict is ONE bound - the one excluding the threshold - not the interval and not both ends. Naming the interval restores the ambiguity M2 exists to remove.
		The point estimate is exactly what the rule forbids quoting as the verdict. Rendering the concept with an estimate word inverts it.
		A Roman terminus was sacred and immovable. A confidence bound is neither, and moves with the evidence.
		McGregor's sense 2 is 'bounds, limit (of morality or convention)'. Reads as propriety, not as a statistical bound.

Back-translation test — status: NOT RUN: no reader has been asked

SEVERITY — severity

Store: TRUE · Domain: method · Confidence: HIGH on the attestations and HIGH on the diagnosis that all of them are wrong for the concept. This entry's value is its forbidden list, not its renderings: Latin and Spanish are confirmed traps and parīkṣā names the test rather than the property. A coinage is likely required. · Reviewer: NONE - PENDING EXPERT REVIEW

Mayo's requirement that a passing test counts as evidence only if it could have failed, made operable as the single question: if the claim were false, what is the probability this exact procedure would still have passed it?

register	form	literal gloss	status
sanskrit	परीक्षा (parīkṣā)	inspection, investigation, examination, test, trial by ordeal	CITED
latin	sevēritās	seriousness, gravity, sternness, strictness	CITED
english	severity	Mayo's requirement that a passing test counts as evidence only if it could have failed, made operable as the single question: if the claim were false, what is the probability this exact procedure would still have passed it?	—
hindi	परीक्षा ()	inspection, investigation, examination; experiment; test, trial; trial by ordeal	CITED
spanish	—	—	NOT-ATTEMPTED

Forbidden mistranslations

register	form	why it is wrong
		L&S s.v. saevĭtĭa: 'a raging, rage, fierceness, ferocity ... harshness, savageness, cruelty, barbarity, severity'. It glosses AS 'severity' in English while meaning savagery, and it is a near-homophone of severitas. The single most available wrong word here.
		RAE: 'Cualidad de severo'; severo = 'Riguroso, áspero, duro en el trato o el castigo'. Says harshness of treatment or punishment. A severe test in Mayo's sense is not harsh and need not be strict; it merely could have failed.
		Relocates the property from the procedure to the person. A stern examiner administering a test that cannot fail has severity zero; a mild one administering a test that would have caught the error has high severity.
		An ordeal's outcome is decided by divine intervention, not by whether the claim is false. It is the exact zero-severity case: the probability of passing is independent of the truth of the claim.
		A difficult test can still be one the claim would pass whether or not it were true. Severity is about the counterfactual failure probability, not about how hard the test is.

Back-translation test — status: NOT RUN: no reader has been asked

STRONG_INFERENCE — strong inference

Store: TRUE · Domain: method · Confidence: MEDIUM on the attestations for anumāna/anumān (solid, but they name inference in general). LOW on fitness: no register carries the exclusion-structure that IS the concept, and the deliberate dropping of 'strong' in the Latin is an unreviewed authorial choice. · Reviewer: NONE - PENDING EXPERT REVIEW

Platt's procedure of devising alternative hypotheses, devising a crucial experiment each of whose possible outcomes EXCLUDES one or more of them, carrying it out to a clean result, and recycling.

register	form	literal gloss	status
sanskrit	अनुमान (anumāna)	the act of inferring or drawing a conclusion from given premises; inference	CITED
latin	illātiō excludēns	the excluding inference	COINED
english	strong inference	Platt's procedure of devising alternative hypotheses, devising a crucial experiment each of whose possible outcomes EXCLUDES one or more of them, carrying it out to a clean result, and recycling.	—
hindi	अनुमान ()	inference; notion, opinion; conjecture, estimate, guess	CITED
spanish	—	—	NOT-ATTEMPTED

Forbidden mistranslations

register	form	why it is wrong
		Says the inference is confident or well-supported. Platt's strength is the exclusion structure of the design. A confident inference with no exclusions is the very thing the procedure is aimed against.
		L&S: 'a carrying or bringing in (post-class.)'; its literal senses include 'mortui', i.e. burial, and 'an impost, duty'. Unfenced it reads as interment or a tax; and it names inference in general, losing the exclusion requirement entirely.
		Names inference as such - a pramāṇa that ESTABLISHES conclusions via vyāpti. Platt's procedure ELIMINATES alternatives. Bare anumāna loses both the plurality of hypotheses and the exclusion.
		The concept is a PROCEDURE with four named steps including recycling. Rendering it as a species of reasoning drops the crucial experiment, the exclusion mapping, and the iteration.

Back-translation test — status: NOT RUN: no reader has been asked

MULTIPLE_WORKING_HYPOTHESES — multiple working hypotheses

Store: TRUE · Domain: method · Confidence: LOW. Latin has no attested word for 'hypothesis' on the route used, three registers are honestly not attempted, and neither the count nor the mechanism-distinctness criterion is rendered anywhere. · Reviewer: NONE - PENDING EXPERT REVIEW

Chamberlin's remedy of plurality rather than neutrality: keep at least two hypotheses always and at least three for anything you intend to claim, each structurally distinct in mechanism rather than in parameter.

register	form	literal gloss	status
sanskrit	—	—	NOT-ATTEMPTED
latin	conjectūrae plūrēs operantēs	several working conjectures	COINED
english	multiple working hypotheses	Chamberlin's remedy of plurality rather than neutrality: keep at least two hypotheses always and at least three for anything you intend to claim, each structurally distinct in mechanism rather than in parameter.	—
hindi	—	—	NOT-ATTEMPTED
spanish	—	—	NOT-ATTEMPTED

Forbidden mistranslations

register	form	why it is wrong
		Chamberlin's remedy is plurality, NOT neutrality. MW s.v. संशय gives 'uncertainty, irresolution, hesitation, doubt' and '(in Nyāya) doubt about the point to be discussed'. Doubt is not a working hypothesis; holding none is not holding several.
		In statistics the alternative hypothesis is the ONE thing opposed to the null. That is precisely the 'one hypothesis plus a null' structure the concept rejects as a strawman.
		In Nyāya, pakṣa is the subject/locus of the inference, and its ordinary MW senses are 'a wing' and 'a side, party, faction'. The faction sense is actively harmful here: it makes rival hypotheses into rival PARTIES, restoring the person-vs-person conflict the method exists to convert into idea-vs-idea.
		A working hypothesis is one you actively WORK, not one you hold loosely. Tentativeness is neutrality wearing a different coat.

Back-translation test — status: NOT RUN: no reader has been asked

REPLICATION — replication

Store: TRUE · Domain: method · Confidence: LOW-MEDIUM. The Latin coinage is the only rendering that carries the discriminating property; Spanish and Hindi are blocked by a confirmed false friend and a confirmed dictionary absence respectively, and are honestly left empty. · Reviewer: NONE - PENDING EXPERT REVIEW

Independent repetition of a measurement by another party, whose presence or absence is exactly what separates OBSERVED-REPLICATED from OBSERVED-SINGLE.

register	form	literal gloss	status
sanskrit	आवृत्ति (āvṛtti)	repetition; recurrence to the same point	CITED
latin	repetītiō aliēna	a repetition belonging to another - a repetition done by someone else	COINED
english	replication	Independent repetition of a measurement by another party, whose presence or absence is exactly what separates OBSERVED-REPLICATED from OBSERVED-SINGLE.	—
hindi	—	—	NOT-ATTEMPTED
spanish	—	—	NOT-ATTEMPTED

Forbidden mistranslations

register	form	why it is wrong
		RAE gives the primary sense as DNA replication (Biol.) and marks the general 'acción de replicar' as desus. - obsolete. The word says molecular biology to a Spanish reader; it does not carry independent scientific replication in current RAE usage.
		L&S: 'A folding or rolling back again (of a book)'; 'In jurid. Lat., a reply, replication'; 'Numeri (opp. multiplicatio), a reduction of a number'. Its 'repetition' sense is attested ONLY in late Latin (Sid. Ep. 1, 5). Dominantly it names a plaintiff's rejoinder in a lawsuit - a reply, not an independent re-measurement - and it never carries 'by another party'.
		Permits the original author repeating their own measurement. That is a repeated run, not a replication, and it does not move OBSERVED-SINGLE to OBSERVED-REPLICATED.
		MW gives 'repeated or permanent exercise, discipline, use, habit, custom' and, in Yoga, 'the effort of the mind to remain in its unmodified condition of purity (sattva)'. It means PRACTICE - repetition for the sake of the practitioner's skill, the opposite of independent verification.

Back-translation test — status: NOT RUN: no reader has been asked

ANOMALY — anomaly

Store: TRUE · Domain: method · Confidence: MEDIUM-HIGH on attestation for Latin, Sanskrit and Spanish; LOW on fitness, since the survival clause that defines the concept is carried by none of them and the Hindi gloss does not even reach 'anomaly'. · Reviewer: NONE - PENDING EXPERT REVIEW

A residual that survives the hunt for instrument error -- stable, isotropic-or-not, and quantitatively characterized -- which is the most valuable object in science and the seed of the trajectory nuisance to residual to characterized anomaly to discriminating test to precision measurement.

register	form	literal gloss	status
sanskrit	व्यभिचार (vyabhicāra)	(in grammar) deviation from or exception to a rule, irregularity, anomaly; going apart or astray, deviating	CITED
latin	anōmalia	(in grammar) irregularity, anomaly	CITED
english	anomaly	A residual that survives the hunt for instrument error -- stable, isotropic-or-not, and quantitatively characterized -- which is the most valuable object in science and the seed of the trajectory nuisance to residual to characterized anomaly to discriminating test to precision measurement.	—
hindi	विसंगति ()	incoherence; disruption	PENDING-CITATION
spanish	anomalía	deviation or discrepancy from a rule or a usage; defect of form or functioning	CITED

Forbidden mistranslations

register	form	why it is wrong
		MW: 'trespass, transgression, crime, vice, sin (esp. infidelity of a wife), Mn.; MBh.' In ordinary usage the word means adultery. Only the (in gram.) and (in phil.) senses are wanted, and they are minority senses.
		RAE sense 3 makes it an angle fixing a body's position in its elliptical orbit - a routine orbital coordinate carrying no suggestion of irregularity at all. In an astrophysical context the word may be read as a parameter.
		L&S attests it only 'in gram.' from Varro, where it is one side of the ancient dispute about whether language follows rule or usage. The Latin word takes a position in a linguistic quarrel; it does not name an empirical residual.
		Exactly inverts the concept. An anomaly is what remains AFTER error and defect have been hunted and excluded. Calling it a defect assigns it to the instrument - the hypothesis the survival clause has already had to defeat.
		An anomaly you cannot describe quantitatively is a rumour. 'Stable, isotropic-or-not, quantitatively characterized' are admission conditions, not decorations.

Back-translation test — status: NOT RUN: no reader has been asked

POSITIVE_CONTROL — positive control

Store: TRUE · Domain: method · Confidence: LOW on renderings - only Latin attempted, as a coinage with one uncited member - and HIGH on the anupalabdhi opposition, which is the entry's real content and is drawn from a quoted MW gloss. · Reviewer: NONE - PENDING EXPERT REVIEW

A run proving the instrument can detect the effect it is looking for, without which a negative result is a hypothesis rather than a finding.

register	form	literal gloss	status
sanskrit	—	—	NOT-ATTEMPTED
latin	probātiō īnstrūmentī	the trial/testing of the instrument	COINED
english	positive control	A run proving the instrument can detect the effect it is looking for, without which a negative result is a hypothesis rather than a finding.	—
hindi	—	—	NOT-ATTEMPTED
spanish	—	—	NOT-ATTEMPTED

Forbidden mistranslations

register	form	why it is wrong
		Says the instrument is being governed or restrained. A positive control governs nothing; it demonstrates detection capability. MW s.v. नियम gives 'restraining, checking, holding back, preventing, controlling' - exactly the wrong sense, and niyama is in any case needed near PRE_REGISTRATION.
		MW: 'non-perception, non-recognition'. It names the doctrine that absence can be KNOWN by non-perception - the inference the positive control exists to forbid. Rendering the guard with the name of the error it guards against is the worst available inversion in this entry.
		A control group is a comparison arm within an experiment. A positive control is a run with a KNOWN answer used to validate the instrument itself. It also collides with BASELINE, a separate concept in this file.
		Without the positive control a negative result is a HYPOTHESIS, not a finding. That downgrade is the concept's entire output and must survive translation.

Back-translation test — status: NOT RUN: no reader has been asked

PROVENANCE — provenance

Store: TRUE · Domain: method · Confidence: MEDIUM on attestations (mūla, mūl, procedencia, index and origo are all solidly quoted); LOW-MEDIUM on fitness, since every register names the origin rather than the maintained record. · Reviewer: NONE - PENDING EXPERT REVIEW

The record of what was measured, on what instrument, by whom, when, and through which cable, kept well enough that the loose connector can be found.

register	form	literal gloss	status
sanskrit	मूल (mūla)	root; basis, foundation, cause, origin; an original text (as opposed to the commentary or gloss)	CITED
latin	index orīginis	the register/pointer of the origin	COINED
english	provenance	The record of what was measured, on what instrument, by whom, when, and through which cable, kept well enough that the loose connector can be found.	—
hindi	मूल ()	root; source, origin; cause; basis, foundation; original (text, or language)	CITED
spanish	procedencia	origin, the beginning from which something arises or derives	CITED

Forbidden mistranslations

register	form	why it is wrong
		L&S: 'Esp., in a bad sense, an informer, betrayer, spy: "Catilinam vallatum indicibus atque sicariis," Cic. Mur. 24, 49.' A provenance record is not a denunciation, and index alone can name a person who informs on someone.
		Names where the thing came from, not the maintained RECORD of it. Drops the operational standard, which is the only part with a test: could the loose connector be found?
		RAE sense 2 is the point of departure of a ship, train or plane. In a logistics register it reads as a departure point on a timetable.
		Those classes record gaps in this corpus's own homework about its citations. This concept records the measurement chain of an observation. One word covering both would merge a claim about the corpus with a claim about an instrument.
		MW and McGregor both record मूल as a lunar asterism (McGregor: 'a lunar asterism (considered unlucky): the stars in the tail of Scorpio') and as 'capital, principal' as opposed to interest. Unfenced it may read as an inauspicious constellation or as a sum of money.

Back-translation test — status: NOT RUN: no reader has been asked

CAVITY_PRINCIPLE — the cavity principle

Store: TRUE · Domain: method · Confidence: LOW, and deliberately so. This is the least translatable concept in the file: its English name is a metaphor borrowed from statistical physics, and every register that renders the metaphor renders nothing. The recommendation on the evidence is to leave it untranslated and gloss the rule. · Reviewer: NONE - PENDING EXPERT REVIEW

M22's rule that a hierarchy level must never treat an upstream prior as fresh evidence, but must divide it out.

register	form	literal gloss	status
sanskrit	—	—	NOT-ATTEMPTED
latin	rēgula cavī	the rule of the hollow/cavity	COINED
english	the cavity principle	M22's rule that a hierarchy level must never treat an upstream prior as fresh evidence, but must divide it out.	—
hindi	—	—	NOT-ATTEMPTED
spanish	—	—	NOT-ATTEMPTED

Forbidden mistranslations

register	form	why it is wrong
		The concept's own note forbids it explicitly: cavity here is a term of art from cavity methods, not an anatomical or dental sense. A dental caries word is pure noise.
		śūnya imports Madhyamaka emptiness - an entire Buddhist metaphysics - onto a bookkeeping rule about not double-counting a prior. The most extravagant possible category error in this file.
		'Cavity' communicates nothing on its own. The translatable object is 'divide out the upstream prior; never treat it as fresh evidence'. A faithful rendering of the hole is an unfaithful rendering of the concept.

Back-translation test — status: NOT RUN: no reader has been asked