Docs / Phonology / OptimalityTheory / HarmonicSerialism

Harmonic Serialism

import Linglib.Phonology.OptimalityTheory.HarmonicSerialism · namespace OptimalityTheory

McCarthy 2000 McCarthy 2010 Pruitt 2023 Lamont 2022a Lamont 2022b

Harmonic Serialism: iterate a one-step gen and a durable constraint ranking to a fixed point. HSDerivation bundles the two; the serial search itself is iterateGen, Function.iterate of a one-round map hsStep on Option C whose fixed points are exactly convergence and failure — durability of both is mathlib's IsFixedPt.iterate; the Nat fuel and the pick tie-breaker make a possibly-non-terminating, possibly-tied search total and computable. One-step EVAL reuses the parallel Tableau machinery; directional EVAL (Lamont 2022b; Eisner 2000) needs nothing extra — a directional constraint like *FLOAT^→ is a position-indexed scalar block (Tone.starFloatBlock) spliced into the ranking under the canonical lex order.

Main definitions

Main results

Theoretical scope

Per Pruitt 2023 (Table 1), HS does not solve counterfeeding underapplication and only patches counterbleeding with extra GEN assumptions; this file makes no "serial OT solves opacity" claim. McPherson & Lamont 2026 is the motivating positive case for the directional variant. Architecturally distinct alternatives in linglib: Phonology/Subregular/LocalRewrite.lean (extrinsic rule ordering) and Phonology/OptimalityTheory/Stratal.lean (cyclic ranking). HS and Stratal OT are sister specializations of parallel OT, not nested: HS holds the ranking constant and iterates the candidate; Stratal OT varies the ranking and chains the candidate (Stratal HS, combining both axes, is deferred — cf. Pruitt 2023 §2.4).

HS derivations

A Harmonic Serialism derivation: a one-step candidate-generation function (per McCarthy 2008b's restricted GEN) and a durable ranking. The inner Tableau width is ranking.length.

structure HSDerivationL169–176
(C : Type u_1) [DecidableEq C] :
Type u_1
  • gen : C → Finset C

    The one-step candidate-generation function.

  • ranking : List (Constraint C)

    The constraint ranking, durable across rounds (head = highest).

Used by HSDerivation.stepOptimum, HSDerivation.stepOptimum_of_empty_gen, HSDerivation.Converged, HSDerivation.instDecidableConverged, HSDerivation.isFixedPt_singleton_iff_converged, HSDerivation.converged_of_singleton_gen, HSDerivation.derive

Used in Lamont2022c, McPhersonLamont2026

The optimal set of one GEN/EVAL round: the winners of the tableau of gen c under the durable ranking.

def HSDerivation.stepOptimumL182–184
{C : Type u_1} [DecidableEq C] (D : HSDerivation C) (c : C) :
Finset C

Built from

Used by HSDerivation.stepOptimum_of_empty_gen, HSDerivation.Converged, HSDerivation.instDecidableConverged, HSDerivation.isFixedPt_singleton_iff_converged, HSDerivation.derive

Used in Lamont2022c, McPhersonLamont2026

L186–187

Convergence

A form has converged iff its one-round optimal set is the singleton of the faithful candidate — the canonical HS halting condition (McCarthy 2010).

def HSDerivation.ConvergedL191–193
{C : Type u_1} [DecidableEq C] (D : HSDerivation C) (c : C) :

Built from ,

Used by HSDerivation.instDecidableConverged, HSDerivation.isFixedPt_singleton_iff_converged, HSDerivation.converged_of_singleton_gen

Used in Lamont2022c, McPhersonLamont2026

Converged is Function.IsFixedPt of the powerset step s ↦ ⋃ x ∈ s, stepOptimum x at the singleton {c} — by Finset.singleton_biUnion, with no intermediate combinator. Simp-normalized toward Converged.

@[simp] theorem HSDerivation.isFixedPt_singleton_iff_convergedL197–203
{C : Type u_1} [DecidableEq C] (D : HSDerivation C) (c : C) :
Function.IsFixedPt (fun (s : Finset C) => s.biUnion D.stepOptimum) {c} ↔ D.Converged c

Built from , ,