Wang (Ruoan) 2023: Honorifics without [HON] #
@cite{wang-r-2023}
Wang, R. (2023). Honorifics without [HON]. Natural Language & Linguistic Theory, 41(4), 1287--1347.
Core Insight #
The cross-linguistic pattern of honorific pronoun recruitment — only PL, 3rd, INDEF are recruited; never SG, 1st/2nd, DEF — falls out from the interaction of phi-feature presuppositions (@cite{sauerland-2003}, @cite{harbour-2016}) with a pragmatic maxim called the Taboo of Directness (ToD).
Key Result #
ToD reverses Maximize Presupposition (@cite{heim-1991}): where MP!
prefers the strongest available presupposition (= most marked form),
ToD prefers the weakest (= least marked = semantically unmarked). The
semantically unmarked values — plural, 3rd person, indefinite — are
precisely those at PrivativePair.minimal (specLevel 0), with vacuous
presuppositions.
No dedicated [HON] feature is needed. The attested patterns are derived
from the same PrivativePair structure that governs ordinary phi-feature
semantics, plus a single pragmatic constraint (ToD).
Architecture #
This file connects three layers:
Features.PrivativePair: the algebraic structure (specLevel ordering)Theories.Semantics.Presupposition.PhiFeatures: presuppositional denotations, semantic markedness, and presuppositional strength orderingCore.Constraint.OT: constraint evaluation and factorial typology
Sections #
- Typological data: the three attested honorific strategies
- ToD and MP! as OT constraints (derived from
presupStrength) - Binary case: ToD >> MP! derives unmarked recruitment
- Ternary case: Strong/Weak ToD for articulated number systems
- [iHON] eliminability — bridge to @cite{alok-bhalla-2026}
- Bridges to
Honorifics.leanand phi-feature denotations - General structural theorem: ToD >> MP! selects unmarked for ANY candidate set
- HonLevel ↔ PrivativePair bridge —
PhiFeatures HonLevelinstance
§1: Typological Data #
Three honorific strategies are attested cross-linguistically — each recruits the semantically unmarked value of a phi-feature domain:
| Domain | Strategy | Unmarked value | Examples |
|---|---|---|---|
| Number | Plural pronoun | PL (specLevel 0) | French, Slovenian |
| Person | 3rd person | 3rd (specLevel 0) | Ainu, Malay |
| Definiteness | Indefinite | INDEF (specLevel 0) | Ainu (DP strategy) |
No language recruits a semantically marked value (SG, 1st/2nd, DEF) for honorification. @cite{wang-r-2023} derives this from the ToD.
The three attested honorific recruitment strategies. Each targets a different phi-feature domain but always recruits the minimal cell (specLevel 0) within that domain.
- plural : HonStrategy
- thirdPerson : HonStrategy
- indefinite : HonStrategy
Instances For
Equations
- Wang2023.instDecidableEqHonStrategy x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- Wang2023.instReprHonStrategy = { reprPrec := Wang2023.instReprHonStrategy.repr }
Map each strategy to the PrivativePair cell that is recruited.
The key empirical generalization: all three strategies map to the
same cell — .minimal (specLevel 0).
Equations
Instances For
All attested strategies recruit the minimal (unmarked) cell. This is the empirical generalization that the ToD analysis derives.
Corollary: all attested strategies recruit a semantically unmarked
value. Follows from all_strategies_use_minimal + minimal_is_unmarked.
A language's honorific profile: which strategies it uses.
- language : String
- strategies : List HonStrategy
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- Wang2023.instReprHonProfile = { reprPrec := Wang2023.instReprHonProfile.repr }
Equations
- Wang2023.french = { language := "French", strategies := [Wang2023.HonStrategy.plural] }
Instances For
Equations
- Wang2023.italian = { language := "Italian", strategies := [Wang2023.HonStrategy.plural, Wang2023.HonStrategy.thirdPerson] }
Instances For
Equations
- Wang2023.german = { language := "German", strategies := [Wang2023.HonStrategy.plural, Wang2023.HonStrategy.thirdPerson] }
Instances For
Equations
- Wang2023.ainu = { language := "Ainu", strategies := [Wang2023.HonStrategy.thirdPerson, Wang2023.HonStrategy.indefinite] }
Instances For
Equations
- Wang2023.slovenian = { language := "Slovenian", strategies := [Wang2023.HonStrategy.plural] }
Instances For
Equations
Instances For
Every language in the typological data uses only unmarked strategies.
§2: Taboo of Directness (ToD) and Maximize Presupposition (MP!) #
ToD: In respect contexts, avoid the form with the strongest
presupposition. Violation count = presupStrength (= specLevel).
MP!: Use the form with the strongest satisfied presupposition.
Violation count = max presupStrength − presupStrength.
ToD and MP! are antagonistic: for any two distinct well-formed cells, they prefer opposite directions. This is the structural heart of @cite{wang-r-2023}'s analysis.
The constraints are defined in terms of presupStrength from
Theories.Semantics.Presupposition.PhiFeatures, not reimplemented —
ToD IS the presuppositional strength ordering.
ToD constraint: penalizes presuppositional strength.
Defined as presupStrength — ToD literally IS the strength
ordering from PhiFeatures, applied as an OT penalty.
Equations
- Wang2023.todConstraint = { name := "ToD", family := Core.Constraint.OT.ConstraintFamily.markedness, eval := Semantics.Presupposition.PhiFeatures.presupStrength }
Instances For
MP! constraint: penalizes failure to maximize presupposition.
Violation count = maxSpecLevel − presupStrength.
PrivativePair.spec_maximal proves maxSpecLevel = 2.
Equations
- One or more equations did not get rendered due to their size.
Instances For
ToD's evaluation IS presupStrength — not a reimplementation.
ToD prefers exactly the presuppositionally weaker cell.
This is the bridge between OT evaluation and the presupWeakerThan
ordering from PhiFeatures: fewer ToD violations ↔ weaker presupposition.
ToD and MP! impose opposite orderings on well-formed cells: ToD prefers c₁ iff MP! prefers c₂.
ToD prefers the minimal (unmarked) cell: it has 0 violations.
MP! prefers the maximal (most marked) cell: it has 0 violations.
mpConstraint is an instance of the general phiMP from
MaximizePresupposition: same eval function, same name, same family.
This connects Wang2023's domain-specific MP! to the general theory.
todConstraint.eval equals markednessPenalty presupStrength.eval.
ToD is an instance of the general markedness penalty from
MaximizePresupposition.
tod_reverses_mp is a corollary of the general
mp_reverses_markedness theorem from MaximizePresupposition.
§3: Binary Phi-Feature Domains #
For binary phi-feature contrasts (SG/PL, 1st/3rd, DEF/INDEF), the candidate set is {maximal, minimal}. Under ToD >> MP!, the optimal candidate is the minimal (unmarked) cell — deriving the universal recruitment of unmarked values for honorification.
Binary candidate set: {maximal, minimal}.
Equations
Instances For
Core prediction: ToD >> MP! selects the minimal (unmarked) cell. Respect contexts recruit the semantically unmarked value.
This is the central theorem: the recruitment generalization
(all_strategies_use_minimal) is DERIVED, not stipulated.
It follows from the interaction of two independently motivated
constraints (ToD from politeness, MP! from presupposition theory)
evaluated over the PrivativePair structure from @cite{harbour-2016}.
Converse: MP! >> ToD selects the maximal (marked) cell. Non-respect speech uses the strongest presupposition. This is the standard Maximize Presupposition from @cite{heim-1991}.
The optimal candidate under ToD >> MP! is semantically unmarked.
Factorial typology: the binary ToD/MP! system predicts exactly 2 language types — honorific (unmarked) vs normal (marked).
§4: Articulated Number (SG/DU/PL) #
Languages with a three-way number distinction (singular/dual/plural) require two ToD strengths:
- SToD (Strong ToD): penalizes ALL marked candidates (specLevel > 0).
Identical to
todConstraint— samepresupStrengthpenalty. - WToD (Weak ToD): penalizes only the MOST marked candidate (specLevel = 2). Tolerates intermediate marking.
The factorial typology over {SToD, WToD, MP!} with candidates {maximal, intermediate, minimal} derives exactly 3 patterns:
| Ranking | Optimal | Interpretation |
|---|---|---|
| MP! dominant | maximal (SG) | Normal speech |
| WToD >> MP! >> SToD | intermediate (DU) | Moderate respect |
| SToD dominant | minimal (PL) | Maximal respect (French) |
Ternary candidate set: {maximal, intermediate, minimal}.
Equations
Instances For
Weak ToD: penalizes only the most marked form (specLevel = max).
Tolerates intermediate marking, unlike todConstraint which
penalizes all marked forms proportionally.
Equations
- One or more equations did not get rendered due to their size.
Instances For
SToD has the same eval function as todConstraint: both penalize
by presupStrength. The ternary case uses todConstraint directly
rather than defining a separate stodConstraint.
WToD >> MP! >> SToD selects the intermediate (dual) cell. Moderate respect in articulated number systems.
SToD >> MP! >> WToD selects the minimal (plural) cell. Maximal respect (French/Slovenian-type pattern).
MP! >> SToD >> WToD selects the maximal (singular) cell. Normal non-honorific speech.
Factorial typology: {SToD (= todConstraint), WToD, MP!} with 3 candidates predicts exactly 3 language types.
No unattested ternary pattern: every constraint permutation selects one of the three canonical cells.
§5: [iHON] is Redundant #
@cite{alok-bhalla-2026} posits a dedicated [iHON] feature in the syntax
(formalized in Minimalist.Features.HonLevel). @cite{wang-r-2023}
argues this is unnecessary: the honorific recruitment pattern falls out
from phiPresup + ToD, without any feature beyond standard phi-features.
The key argument: [iHON] + impoverishment rules must stipulate which phi-values are recruited (always the unmarked one). ToD + phi-feature presuppositions derive this — the unmarked value wins because it has the weakest presupposition, and ToD selects the weakest.
Note: @cite{alok-bhalla-2026}'s analysis of allocutive Agree (probe locus, embeddability) is orthogonal — [iHON] may play a role in the agreement mechanism even if it is unnecessary for recruitment.
The phi-feature presuppositional framework + ToD derives all attested recruitment patterns without [iHON]:
- ToD >> MP! selects the minimal cell
- The minimal cell is semantically unmarked
- All attested strategies target the minimal cell
§6: Bridges #
Per-domain bridges #
Each recruitment strategy targets a specific phi-feature domain.
The recruited cell (.minimal) corresponds to a specific PrProp
denotation from PhiFeatures: plSem (number), thirdSem (person),
or indefSem (definiteness). All three are phiPresup at the minimal
cell, which has a vacuous presupposition — this is WHY ToD selects them.
Allocutive data bridges #
The allocutive data in Honorifics.lean tracks hasTV (T/V pronoun
distinction = plural recruitment) and has3PHon (3rd-person honorifics
= person recruitment). Both correspond to the .minimal cell in their
respective phi-feature domains.
The plural recruitment strategy targets the minimal NUMBER cell,
which is plSem — the PrProp with vacuous presupposition.
pl_is_minimal_cell (PhiFeatures) proves pluralF maps to .minimal.
The 3rd-person strategy targets the minimal PERSON cell,
which is thirdSem — the PrProp with vacuous presupposition.
Both strategies target cells whose presuppositional denotations
are vacuously defined — this is the semantic reason ToD selects them.
Proved via unmarked_vacuous_presup from PhiFeatures.
The hasTV field in AllocDatum tracks whether a language uses
the plural (= minimal number cell) recruitment strategy.
All 9 allocutive languages have T/V. Cross-reference: Honorifics.all_have_tv.
Languages with has3PHon = true use the minimal PERSON cell.
The languages are: Magahi, Korean, Japanese, Tamil, Hindi, Maithili, Punjabi.
Dual-domain languages (hasTV ∧ has3PHon) recruit the minimal cell independently in both the number and person domains.
§7: General Structural Theorem #
The binary-case theorem (tod_mp_selects_minimal) uses native_decide over
a fixed 2-element candidate set. Here we prove the general result: for
ANY non-empty set of well-formed PrivativePair candidates that includes
.minimal, ToD >> MP! selects .minimal as the unique winner.
This is the structural backbone of @cite{wang-r-2023}'s analysis: the recruitment of unmarked values is not an accident of the binary case — it holds for arbitrary candidate sets. The proof is purely algebraic:
optimal_zero_first: if any candidate has 0 violations on the top constraint, all optimal candidates must too- The only well-formed cell with
presupStrength = 0is.minimal .minimal's profile[0, maxSpec]lexicographically dominates all other profiles
Every optimal candidate under ToD >> MP! is .minimal. The proof:
optimal_zero_first gives todConstraint.eval c = 0, i.e.
presupStrength c = 0. A case split on PrivativePair's fields
shows .minimal is the only well-formed cell with specLevel 0.
.minimal is in the optimal set: its profile [0, maxSpec] is
lexicographically ≤ every profile [k, maxSpec - k] for k : Nat.
General ToD >> MP! Theorem: for any set of well-formed candidates
containing .minimal, the optimal set under ToD >> MP! is exactly
{.minimal}. This is the strongest formulation of @cite{wang-r-2023}'s
core result — the recruitment of semantically unmarked values is a
necessary consequence of presuppositional competition under ToD
dominance, regardless of candidate set size or composition.
§8: HonLevel ↔ PrivativePair Bridge #
@cite{alok-bhalla-2026}'s HonLevel (nh/h/hh) is isomorphic to
PrivativePair (minimal/intermediate/maximal). This PhiFeatures
instance makes the correspondence structural: HonLevel inherits
specLevel, wellFormed, no_four_way, and all presuppositional
machinery from PrivativePair by construction.
The mapping:
nh↔.minimal(specLevel 0 — unmarked, weakest presupposition)h↔.intermediate(specLevel 1)hh↔.maximal(specLevel 2 — most marked, strongest presupposition)
This makes ihon_redundant_for_recruitment structurally precise:
HonLevel values are literally PrivativePair cells, so the
tod_mp_general result from §7 applies directly to HonLevel
candidates.
Equations
- One or more equations did not get rendered due to their size.
All HonLevel values are well-formed as PrivativePair cells.
specLevel values: nh = 0, h = 1, hh = 2.
nh maps to .minimal — the cell ToD selects.
hh maps to .maximal — the cell MP! selects.
The HonLevel → specLevel map is injective: distinct levels have
distinct specification levels. This means the 3-way honorific distinction
saturates the PrivativePair structure — no finer distinctions are possible.
HonLevel is fully discriminatory: distinct levels ↔ distinct specLevels.
The forward direction (≠ → specLevel ≠) is the contrapositive of injectivity;
the reverse (specLevel ≠ → ≠) is trivial.
Structural [iHON] eliminability. The PhiFeatures HonLevel
instance means tod_mp_general applies directly to HonLevel
candidates (via toPair): ToD >> MP! selects nh (= .minimal)
as the unique winner whenever nh is among the candidates.
Combined with ihon_redundant_for_recruitment, this shows [iHON]
is not just empirically redundant but structurally so: HonLevel
IS PrivativePair, and PrivativePair + ToD already determines
the recruitment pattern.