Marco & Rasin (2026): Optimal Paradigms — A Challenge from JTA #
@cite{marco-rasin-2026} @cite{mccarthy-2005}
Formalizes the argument that Optimal Paradigms (OP) cannot simultaneously predict the distribution of schwa in verbs, nouns, and adjectives in Judeo-Tripolitanian Arabic (JTA). The core paradox:
- Verbs: schwa always between C₂ and C₃ (CCəC), regardless of sonority
- Adjectives: same as verbs — CCəC regardless of sonority
- Nouns: sonority-sensitive — CəCC when C₂ > C₃, CCəC otherwise
Adjectives share phonological behavior with verbs but paradigm structure with nouns (only V-initial suffixes). OP predicts phonological behavior should track paradigm structure, so it wrongly predicts adjectives should behave like nouns.
Modeling note #
In OP, candidates are whole paradigms where each member independently chooses a stem shape. Markedness constraints (*CCC, *ə]σ) determine the shapes of suffixed members: C-initial suffixed forms must be CCəC (to avoid tri-consonantal clusters), V-initial suffixed forms must be CəCC (to avoid schwa in an open syllable). Only the suffix-less bare form is free — its shape is determined by OP-MAX-V (paradigm uniformity).
For verbs, we enumerate the 4 candidates from tableau (10) in the paper.
For adjectives, we enumerate all 2³ = 8 possible paradigms exhaustively.
The main result (adj_always_initial) proves that the uniform CəCC
paradigm has 0 violations on all four constraints and thus wins under
every ranking — contradicting the attested CCəC pattern.
The paper considers two scenarios for adjectives: with a paradigm
(tableau (19)) and without (tableau (20)). Both fail — op_wrong_adjectives
covers (19) and op_wrong_adj_bare covers (20), formalizing the paper's
argument that OP fails "regardless of what constitutes a paradigm."
Equations
- MarcoRasin2026.instDecidableEqSchwaPos x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- MarcoRasin2026.instReprSchwaPos = { reprPrec := MarcoRasin2026.instReprSchwaPos.repr }
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
Equations
- MarcoRasin2026.instBEqSchwaPos.beq x✝ y✝ = (x✝.ctorIdx == y✝.ctorIdx)
Instances For
Equations
Suffix onset type, determining the phonological environment.
- cInit : SuffixOnset
- vInit : SuffixOnset
- bare : SuffixOnset
Instances For
Equations
- MarcoRasin2026.instDecidableEqSuffixOnset x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- MarcoRasin2026.instReprSuffixOnset = { reprPrec := MarcoRasin2026.instReprSuffixOnset.repr }
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
Equations
- MarcoRasin2026.instBEqSuffixOnset.beq x✝ y✝ = (x✝.ctorIdx == y✝.ctorIdx)
Instances For
A single paradigm member: schwa position + suffix type.
- schwa : SchwaPos
- suffix : SuffixOnset
Instances For
Equations
- MarcoRasin2026.instDecidableEqJTAForm.decEq { schwa := a, suffix := a_1 } { schwa := b, suffix := b_1 } = if h : a = b then h ▸ if h : a_1 = b_1 then h ▸ isTrue ⋯ else isFalse ⋯ else isFalse ⋯
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- MarcoRasin2026.instReprJTAForm = { reprPrec := MarcoRasin2026.instReprJTAForm.repr }
Equations
Equations
- MarcoRasin2026.instBEqJTAForm.beq { schwa := a, suffix := a_1 } { schwa := b, suffix := b_1 } = (a == b && a_1 == b_1)
- MarcoRasin2026.instBEqJTAForm.beq x✝¹ x✝ = false
Instances For
*CCC: assign * for every sequence of three consonants. Violated by CəCC (initial) stems before C-initial suffixes: C₁əC₂C₃ + C → C₂C₃C = tri-consonantal cluster.
Equations
- One or more equations did not get rendered due to their size.
Instances For
*ə]σ: assign * for every schwa in an open syllable. Violated by CCəC (medial) stems before V-initial suffixes: C₁C₂ə.C₃V → schwa in open syllable.
Equations
- One or more equations did not get rendered due to their size.
Instances For
SONCON: assign * for a CCəC (medial) form when C₂ > C₃ in sonority.
Parametrized over the sonority ranks of C₂ and C₃, using the
LinearOrder SonorityRank instance from Syllable.Defs.
Equations
- One or more equations did not get rendered due to their size.
Instances For
OP-MAX-V: assign * for each ordered pair of paradigm members with different schwa positions. Each positional mismatch = 1 violation. For n₁ medial and n₂ initial members: n₁·n₂ + n₂·n₁ = 2·n₁·n₂ total violations (matching @cite{mccarthy-2005}'s counting).
Equations
- MarcoRasin2026.opMaxV = Phonology.ParadigmUniformity.mkOPMaxV fun (f1 f2 : MarcoRasin2026.JTAForm) => if (f1.schwa != f2.schwa) = true then 1 else 0
Instances For
JTA verbal suffixes: bare + 4 C-initial + 2 V-initial = 7 members.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Helper: build a paradigm from a list of schwa positions and suffix list.
Equations
- One or more equations did not get rendered due to their size.
Instances For
(10a) ☞ Bare=medial: 5 medial, 2 initial. OP-MAX-V = 20.
Equations
- One or more equations did not get rendered due to their size.
Instances For
(10b) Bare=initial: 4 medial, 3 initial. OP-MAX-V = 24.
Equations
- One or more equations did not get rendered due to their size.
Instances For
(10c) Uniform medial: *ə]σ violated twice.
Equations
- One or more equations did not get rendered due to their size.
Instances For
(10d) Uniform initial: *CCC violated four times.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
Instances For
Nouns: bare singular only (1 member).
Equations
Instances For
Equations
Instances For
Equations
Instances For
Adjective suffixes: bare (MSG) + 2 V-initial (-a FSG, -in PL).
Equations
Instances For
All 2³ = 8 possible adjective paradigms. Exhaustive enumeration ensures the negative result holds for ALL candidates, not just selected ones.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The attested adjective paradigm: bare=medial, V-init=initial.
Equations
Instances For
The OP-predicted paradigm: uniform initial.
Equations
Instances For
McCarthy's ranking: *ə]σ, *CCC ≫ OP-MAX-V ≫ SONCON (for C₂ > C₃).
Equations
- One or more equations did not get rendered due to their size.
Instances For
OP correctly selects the attested verbal paradigm (10a): bare=medial. Majority Rules in action — the bare form aligns with the C-initial majority (5 medial vs 2 initial members).
OP correctly selects CəCC for nouns (when C₂ > C₃). Single-member paradigm, so OP-MAX-V is vacuous; SONCON decides.
OP wrongly selects uniform-initial for adjectives (tableau (19)). Adjectives have only V-initial suffixes (like nouns), so OP-MAX-V cannot force medial schwa in the bare form. SONCON then favors initial.
The attested adjective paradigm (bare=CCəC) is NOT selected by OP.
The four constraints used in McCarthy's OP analysis.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The uniform-initial adjective paradigm has zero violations on all four OP constraints. This is the structural reason it wins under every ranking — not a coincidence of McCarthy's particular ranking.
- *ə]σ: 0 — initial schwa (CəCC) is never in an open syllable
- *CCC: 0 — no C-initial suffixes in the adjective paradigm
- OP-MAX-V: 0 — uniform paradigm, no positional mismatches
- SONCON: 0 — no medial (CCəC) forms to violate sonority
A candidate with profile [0,0,...,0] is the minimum of any
lexicographic order on Nat-valued profiles (ViolationProfile.zero_le),
so permuting the ranking cannot change the outcome.
Main result. No ranking of the four OP constraints selects the attested adjective paradigm (bare=CCəC).
The uniform-initial paradigm has zero violations on all four
constraints (adjOPpred_zero_viols), making it the lexicographic
minimum under any permutation (adjOPpred_always_optimal). Moreover,
it is the unique winner — no other candidate also has a zero
profile (verified by decide).
Since [0,0,0,0] is the lexicographic minimum regardless of constraint permutation, uniform-initial wins under every ranking. The attested paradigm (bare=medial, V-init=initial) always loses because it incurs OP-MAX-V and SONCON violations.
This formalizes @cite{marco-rasin-2026}'s central challenge: OP cannot derive category-distinct phonological behavior when two categories share paradigm structure but differ phonologically.
If adjectives have no inflectional paradigm (single bare form, like
nouns), SONCON wrongly favors CəCC (initial). This is tableau (20)
in the paper: the single-member adjective case. Combined with
op_wrong_adjectives (tableau (19)), this shows OP fails
regardless of what constitutes a paradigm for JTA adjectives.
The attested adjective form (CCəC = medial) is NOT selected when adjectives are treated as having no paradigm.
Equations
- MarcoRasin2026.instDecidableEqMorphCat x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- MarcoRasin2026.instReprMorphCat = { reprPrec := MarcoRasin2026.instReprMorphCat.repr }
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- MarcoRasin2026.instBEqMorphCat.beq x✝ y✝ = (x✝.ctorIdx == y✝.ctorIdx)
Instances For
Equations
{CCəC}_{Verb, Adj}: template constraint applying to verbs and adjectives but not nouns. Penalizes initial (CəCC) forms in verbs and adjectives. Directly references morphosyntactic categories, violating the TCNP (@cite{bobaljik-2008}).
Equations
- One or more equations did not get rendered due to their size.
Instances For
Category-specific ranking: {CCəC}_{V,A} ≫ SONCON.
Equations
Instances For
Template correctly selects medial (CCəC) for verbs. (Using uniform candidates since template forces all members to medial.)
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Instances For
Template correctly selects initial (CəCC) for nouns (template inactive).
Template correctly selects the attested adjective paradigm — the case where OP fails. The template forces medial (CCəC) for both verb and adjective categories, overriding SONCON.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Each unique-winner tableau lifts to a generic ConstraintSystem via
tableauSystem. For McCarthy's OP ranking this exposes the correct
predictions for verbs and nouns, and the wrong prediction for adjectives —
the empirical content of @cite{marco-rasin-2026}'s argument.
Verbal paradigm under McCarthy's OP ranking.
Equations
Instances For
OP correctly predicts the attested bare-medial verbal paradigm (10a).
Nominal paradigm under McCarthy's OP ranking.
Equations
Instances For
OP correctly predicts CəCC for nouns when C₂ > C₃.
Adjectival paradigm under McCarthy's OP ranking. The probability-1 prediction here is the wrong uniform-initial paradigm — the formal content of @cite{marco-rasin-2026}'s challenge.
Equations
Instances For
OP wrongly assigns probability 1 to the uniform-initial adjective paradigm. The attested CCəC pattern receives probability 0 — the formal sharpening of "OP fails for adjectives."
Loser side: the attested adjective paradigm gets probability 0.
Verbal paradigm under the category-specific template ranking.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Adjectival paradigm under the category-specific template ranking — now correctly predicting the attested medial pattern with probability 1.
Equations
- One or more equations did not get rendered due to their size.