← All conjectures · South Asian text cultures

Ten lost grammars hide in the options

Iron Age & Antiquity Dated c. 600 BCE – 300 BCE South Asia

Status: Falsified

The verdict’s fine print — quoted from the resolution record: “KILLED across the entire reading band -- the best available reading (BROAD, 6/18 = 0.333) reaches well under half the wagered 0.80, so no defensible classification choice rescues clause (2); this is a clean kill on well-measured numbers, not an operationalizat…” Read the full caveats ↓

Status is derived only from the shepherd-authored triage/prediction data above -- community submissions and claims are a separate overlay and can never change it (see the participation panel below).

This is a conjecture imagined by a language model — drawn from its trained weights and held to falsifiability, novelty, and plausibility, not to any one method: it may join two or more fields, or none. It is not an article and not evidence: it sits below the evidence/publication boundary. A quantitative prediction and a named kill-dataset are attached (when registered) so the claim stays falsifiable rather than merely evocative.

Claim (verbatim)

The Aṣṭādhyāyī names ten predecessors — Āpiśali, Kāśakṛtsna, Gārgya, Gālava, Cākravarmaṇa, Bhāradvāja, Śākaṭāyana, Śākalya, Senaka, Sphoṭāyana — and every one of their grammars is lost: an entire pre-Pāṇinian discipline surviving as names in the winner's text. But the names are not decoration. An observation going back to Kielhorn holds that Pāṇini cites a teacher by name precisely where doctrines diverged and both outputs circulated — name-citation as option-licensing, the doxographic footnote of a winner-take-all pedagogy. If that is right, ten grammatical systems were compressed into a residue of licensed doublets, and the residue is countable: rules citing predecessors should overwhelmingly generate optional forms, while the anonymous mass of the grammar is obligatory. Prediction: enumerating every Aṣṭādhyāyī sūtra that cites a predecessor by name and classifying each via the Kāśikāvṛtti as optional (licensing doublet forms) or obligatory, at least 80% of the name-citing rules will be optional, against under 20% for a size-matched random sample of anonymous sūtras (primary clause: the 80% optionality share; the verdict follows it). Exact computation: locate the name-citing sūtras by searching the e-text for the teachers' genitives, then classify each rule from the Kāśikā's gloss. Kill: the GRETIL e-texts of the Aṣṭādhyāyī and the Kāśikāvṛtti, with the register of predecessor citations in George Cardona's Pāṇini: A Survey of Research (1976) as the control list.

Prediction clause (verbatim)

Prediction: enumerating every Aṣṭādhyāyī sūtra that cites a predecessor by name and classifying each via the Kāśikāvṛtti as optional (licensing doublet forms) or obligatory, at least 80% of the name-citing rules will be optional, against under 20% for a size-matched random sample of anonymous sūtras (primary clause: the 80% optionality share; the verdict follows it). Exact computation: locate the name-citing sūtras by searching the e-text for the teachers' genitives, then classify each rule from the Kāśikā's gloss.

Kill-dataset (verbatim)

Kill: the GRETIL e-texts of the Aṣṭādhyāyī and the Kāśikāvṛtti, with the register of predecessor citations in George Cardona's Pāṇini: A Survey of Research (1976) as the control list.

Provenance

Run: Fresh agent generation · model: claude-fable-5

Fresh blind generation by claude-fable-5, 2026-07-16, breadth wave weighting India/South Asia by inferred textual production rather than survival; every item grounded in real works, authors, catalogues, and testimonia, including the real evidence of loss (citing authors, catalogue entries, translation corpora, rediscovery cases); no fabricated citations.

Novelty / leakage triage

anticipated in the literature — this exact test has never been run

The option-licensing reading of Panini's name-citations is a long-standing observation (going back to Kielhorn's era and debated within the tradition itself as pujartha versus vikalpa; surveyed by Cardona and Scharfe), and the register of predecessor-citing sutras is enumerated in the surveys. The quantitative census — classifying every name-citing rule via the Kasika against a size-matched anonymous control sample — has not been run.

Sources cited by the triage

G. Cardona, Panini: A Survey of Research (Mouton, 1976), section on the pre-Paninian grammarians
H. Scharfe, Grammatical Literature (History of Indian Literature V.2, Harrassowitz, 1977)

Predictions

Killed registered 2026-07-17 calibration prediction (parent triage: leaked/adjacent)

Resolution: Killed

Caveats: KILLED across the entire reading band -- the best available reading (BROAD, 6/18 = 0.333) reaches well under half the wagered 0.80, so no defensible classification choice rescues clause (2); this is a clean kill on well-measured numbers, not an operationalization failure. CALIBRATION NOTE: this one is worse than the usual overshoot pattern -- under STRICT the DIRECTION ITSELF REVERSES (name-citing 5.6% vs anonymous 6.7%), and under the literally registered size-matched-sample comparison the broad shares are IDENTICAL (6/18 vs 6/18, zero gap); only the full-population BROAD comparison shows the conjectured direction at all (33.3% vs 23.8%, ~1.4x -- real but modest, nowhere near 'overwhelmingly'). Kielhorn's kernel survives in qualitative miniature: the Kasika itself sorts name-citations into pujartham (honorific -- e.g. 7.3.99, 8.3.20) vs vibhasartham/vikalpartham (option-licensing -- 1.1.16, 6.1.123, 6.1.130, 8.3.19), and exactly one sutra (6.1.92 va supy apisaleh) carries the operator in its own text; so name-citation is SOMETIMES option-marking, but the tradition's own gloss says it is often mere respect. DISCLOSED-NO-WEIGHT: (a) two of the 18 are taddhita/gotra uses of 'Bharadvaja' rather than authority-citations -- dropping them (N=16) gives 1/16 = 0.063 / 6/16 = 0.375, still failing by a wide margin, so the verdict is robust to that ambiguity; (b) Kasakrtsna and Senaka never appear in sutra text while off-list Kasyapa does, suggesting the registered ten-name roster follows a traditional enumeration not fully matching the received text -- immaterial to the verdict since the guard is on located sutras (18 >= 8), not located teachers; (c) STRICT's 'unambiguously governs' disjunct was resolved mechanically to operator-in-sutra-text only, with gloss-stated anuvrtti captured under BROAD -- a judgment call applied identically to both populations, and the reported range brackets both resolutions. Single GRETIL edition (Aryendra Sharma via Minamino), no manuscript collation.

Registered against the GRETIL e-texts of the Astadhyayi and the Kasikavrtti. Claim under test (primary clause): among the Astadhyayi sutras that cite a named predecessor grammarian, AT LEAST 80% license an OPTIONAL rule (a doublet/alternative form), against under 20% for a size-matched random sample of non-name-citing sutras.

Resolution criteria — the registered fine print

Resolution criteria: DATA: GRETIL Astadhyayi + Kasikavrtti e-texts; freeze sha256. Cardona 1976 (Panini: A Survey of Research) register of predecessor citations as the control list of expected name-citing sutras. POPULATION: name-citing sutras = those naming one of Apisali, Kasakrtsna, Gargya, Galava, Cakravarmana, Bharadvaja, Sakatayana, Sakalya, Senaka, Sphotayana (located by the teacher's genitive in the sutra text). CLASSIFY each such sutra OPTIONAL vs OBLIGATORY from the Kasika's gloss (OPTIONAL = the gloss marks an alternative via va / vibhasa / anyatarasyam and cognate optionality terms). PRIMARY = optionality-share among name-citing sutras. COMPARISON = optionality-share in a size-matched random sample of non-name-citing sutras. CLAUSE PRECEDENCE: (1) INCONCLUSIVE_BY_DESIGN if the e-texts cannot be parsed/aligned, OR fewer than 8 name-citing sutras are located, OR the Kasika optionality gloss cannot be classified. (2) SUPPORTED if optionality-share(name-citing) >= 0.80 AND optionality-share(anonymous sample) < 0.20. (3) KILLED otherwise. Report the two shares, the located name-citing sutra list (for audit), and the classification rule; if optionality classification admits more than one reasonable reading report the share range. computed_at postdates registered_at.

Known-priors disclosure — what the registrant already knew

Known priors disclosure: Held: Kielhorn's classic observation that Panini's named predecessors survive as optionality residue, and general knowledge the Astadhyayi + Kasika are in GRETIL. NOT computed: the optionality shares. The 80%/20% thresholds are the conjecture's own.

Method and dataset — how it was measured

Firewalled blind compute (claude-sonnet-5, blind to the 80%/20% thresholds and to direction) fetched, parsed, and classified both GRETIL files with stdlib-only tooling, reporting the registered share range (STRICT/BROAD) plus both anonymous comparisons (full 3,933-sutra population, deterministic; and a seed-42 size-matched sample of 18). The Opus orchestrator independently re-derived the strict count and spot-verified 8.3.19/6.1.123/6.1.130/6.1.127 against the raw gloss. Verdict authored by claude-fable-5 per the registered clause precedence: guard (1) passed (clean parse/alignment, 18 >= 8 located, gloss classified); clause (2) requires P_name >= 0.80 AND P_anon < 0.20 -- P_name is 0.0556 STRICT / 0.3333 BROAD, failing the 0.80 bar under every reading (and BROAD P_anon 0.2377 independently fails the < 0.20 arm); clause (3) therefore fires: KILLED.

Dataset: GRETIL Kasikavrtti jvkasixu.htm (sha256 365e16a2b08ea0dfb28bb5c6c979833d6c8798edd4bf3c33f714b038d795bf89, sutra text + gloss) cross-checked against GRETIL Astadhyayi paninixu.htm (sha256 3eceef025fa631b90ee788db727c9c0d46069bfa22cee0fe8d7a0b71bc3314c1, sutra-only): both files agree exactly on a population of 3,951 PS-marked sutras, every ref with exactly one sutra-text block and one gloss block (verified programmatically). N=18 name-citing sutras located (guard >=8 PASSES); 8 of the 10 registered teachers found in sutra text; Kasakrtsna and Senaka appear in zero sutras (gloss-only occurrences), while Kasyapa -- not on the registered list -- appears in 4 sutras (neutral observation on the fixed input list). Optionality classified mechanically (va/vibhasa/anyatarasyam) under two pre-declared readings applied identically to both groups: STRICT = operator in the sutra's own text; BROAD = operator anywhere in the sutra's Kasika gloss (incl. explicit anuvrtti/vibhasartham statements). Guard clause (1) NOT triggered on any arm.

computed 2026-07-20

Weigh in

No community feedback yet.

New here? Create an account first

Create an account or sign in and your feedback is tied to you — you can track it, get replies, and claim this conjecture so others know you’re working on it. Prefer not to? Just leave your take below as a guest — only the name you type is shown.

Add your take

Posted immediately (spam is removed). Community feedback is never an adjudicated verdict and never changes this conjecture's triage label or status above.