Created: 25-09-2026 · Last updated: 25-09-2026

title: Access structures — printed sort order toc: false

Access structures — what rule produced the printed order?

Wiegand's Zugriffsstruktur is the ordering a reader relies on to find an entry. This page carries the census behind docs/ACCESS_STRUCTURES_SORT_ORDER_CENSUS_24-09-2026.md: for MW, PWG, PW (PWK), AP, WIL, SKD and VCP it recovers the headword order actually printed (the csl-orig record order), tests it against a grid of 72 varṇamālā rules plus two controls, and reports the best-fit rule, its violation rate, the evidence that separates it from its rivals, and every counterexample.

The two versified kośas (ARMH, ABCH) are ordered by concept, not by alphabet, and are measured elsewhere. Sanskrit is displayed in IAST throughout; the SLP1 sort key is kept as a muted secondary column.

Trust Block

At a glance

Best fit per dictionary

A violation is an adjacent pair of headwords that goes down under the rule, counted inside one alphabet run. Displaced is the minimum share of headwords that would have to move for the order to satisfy the rule. The roman control is the same order read as IAST in Latin-alphabet order: at 31–37 % everywhere, it is what shows that every one of these dictionaries follows the varṇamālā.

Factor status — what the order can and cannot decide

Overall rates cannot settle a factor: most adjacent pairs never touch ṃ, ḥ or a geminate after r. Each factor is judged on pairwise contrasts — only the pairs where two of its policies disagree, the other factors held at best fit. Fewer than such pairs is undetermined; a winner at 3 : 1 or better is clear. partial means some alternatives are rejected and others are indistinguishable from the chosen policy.

AP's root nests — and why vowel grades do not explain them

Apte prints derivatives under their verbal root, in derivational rather than alphabetical order, so its rate is a macrostructure effect. H5409 tested whether the residue is derivatives carrying the root vowel in guṇa or vṛddhi grade (√kṛ nesting kāra, karaṇa): the heads_nest view collapses those into the root's position. It removes only an eighth of the disorder, so refutation condition 3 of the methods page fired — the vowel-grade explanation is refuted, and prefix-family grouping is the untested alternative.

Counterexample browser

Every within-run descent of the best-fit rule in each dictionary's primary view, with its L-id, page reference and the pair in IAST. kind is block when the pair sits inside a stretch of ≥ consecutive displaced headwords (a bodily misplaced passage) and isolated otherwise. sensitive_to_anusvara_visarga_policy marks the pairs that some other rule in the grid would order correctly — the pairs that carry the evidence about ṃ and ḥ.

The full rate table

Every dictionary × view × rule combination the census scored — the raw material behind the best-fit choices above.

Method in one paragraph

The sort key is k1, the accentless normalised SLP1 headword; consecutive records sharing a key collapse into one position. Alias records ({{Lbody=N}}, spellings added by the digitisers) are dropped from every view but all; MW and AP additionally give every run-on its own record, so their primary view keeps level-1 heads only, and AP's drops paragraph run-ons and root-nest derivatives on top of that. A supplement or Nachträge run is detected as an alphabet restart — the initial letter goes backwards and the next headwords all sort below the previous — and is reported as a separate run rather than counted as violations. Full method, unknowns and refutation conditions: the methods page.

Dr. Mārcis Gasūns