Presuppositional Semantics of Phi-Features #
@cite{sauerland-2003} @cite{sauerland-2008b} @cite{harbour-2016} @cite{heim-1991} @cite{wang-r-2023}
Phi-features (number, person, definiteness) are presuppositional partial
identity functions on the entity domain, ordered by presuppositional
strength via Features.PrivativePair.specLevel.
The core mathematical object is phiPresup: a single function that maps
each PrivativePair cell to a PrProp, using two predicates (innerP,
outerP) corresponding to the inner and outer privative features. Since
the three well-formed cells have 2, 1, and 0 marked features respectively,
their presuppositions are automatically nested — more marked features =
stronger presupposition = smaller domain.
Domains #
| Domain | innerP | outerP | maximal (2) | intermediate (1) | minimal (0) |
|---|---|---|---|---|---|
| Number | Atom | MinimalGroup | singular | dual | plural |
| Person | speaker ≤ · | speaker ≤ · ∨ addressee ≤ · | 1st | 2nd | 3rd |
| Gender | isInanimate | isFemale | neuter | feminine | masculine |
| Definiteness | familiar/unique | — | definite | — | indefinite |
Semantic Markedness (@cite{wang-r-2023}) #
The semantically unmarked values (plural, 3rd person, indefinite) are precisely those at the minimal cell (specLevel 0) with vacuous presuppositions. These are the values recruited cross-linguistically for honorification — an observation that falls out from the presuppositional framework without stipulation.
Architecture #
This file was extracted from Sauerland2003 to
separate general phi-feature presuppositional theory (which belongs in
Theories/) from Sauerland's specific arguments about number (which
belong in Phenomena/Plurals/Studies/).
Generic presuppositional denotation from a privative feature pair.
Maps each PrivativePair cell to a PrProp using two predicates:
innerP for [±inner] and outerP for [±outer].
| Cell | outer | inner | Presupposition |
|---|---|---|---|
| maximal | + | + | innerP |
| intermediate | + | − | outerP |
| minimal | − | − | vacuous |
Since [+inner] → [+outer] (privative containment), innerP
implies outerP. So maximal's presupposition (innerP) is the
strongest — no need to separately conjoin outerP.
Equations
- Semantics.Presupposition.PhiFeatures.phiPresup innerP outerP { outer := true, inner := true } = { presup := innerP, assertion := fun (x : E) => True }
- Semantics.Presupposition.PhiFeatures.phiPresup innerP outerP { outer := true, inner := false } = { presup := outerP, assertion := fun (x : E) => True }
- Semantics.Presupposition.PhiFeatures.phiPresup innerP outerP { outer := false, inner := inner } = { presup := fun (x : E) => True, assertion := fun (x : E) => True }
Instances For
Feature-Subset Principle, derived from privative geometry.
If innerP → outerP (the containment [+inner] → [+outer]), then
more specified cells have smaller presuppositional domains. This
is the semantic content of PrivativePair.spec_strict_order —
not a stipulation but a consequence of the algebraic structure.
All phiPresup cells have the same (trivial) assertive content.
This is the privative-geometric reason why φ-feature competition
is presuppositional, not at-issue.
⟦Sg⟧: presupposes atomicity. The identity function restricted to atoms — defined only when the referent is an atom.
Equations
- Semantics.Presupposition.PhiFeatures.sgSem E = { presup := Mereology.Atom, assertion := fun (x : E) => True }
Instances For
⟦Pl⟧: no inherent presupposition. The unrestricted identity function. Its distribution is constrained pragmatically by Maximize Presupposition, not by any semantic content.
Equations
- Semantics.Presupposition.PhiFeatures.plSem E = { presup := fun (x : E) => True, assertion := fun (x : E) => True }
Instances For
⟦Du⟧: presupposes minimality (no proper non-atomic subpart). The intermediate cell (specLevel 1).
Equations
- Semantics.Presupposition.PhiFeatures.dualSem minimalP = { presup := minimalP, assertion := fun (x : E) => True }
Instances For
Singular features map to the maximal PrivativePair cell (specLevel 2).
Plural features map to the minimal cell (specLevel 0).
The presuppositional asymmetry tracks specification level: singular (specLevel 2) has content; plural (specLevel 0) is vacuous.
⟦1st⟧: presupposes the referent includes the speaker. Maximal cell [+author, +participant] (specLevel 2).
Equations
- Semantics.Presupposition.PhiFeatures.firstSem speaker = { presup := fun (x : E) => speaker ≤ x, assertion := fun (x : E) => True }
Instances For
⟦2nd⟧: presupposes the referent includes a speech-act participant. Intermediate cell [−author, +participant] (specLevel 1).
Equations
- Semantics.Presupposition.PhiFeatures.secondSem speaker addressee = { presup := fun (x : E) => speaker ≤ x ∨ addressee ≤ x, assertion := fun (x : E) => True }
Instances For
⟦3rd⟧: vacuous presupposition. Minimal cell [−author, −participant] (specLevel 0).
Equations
- Semantics.Presupposition.PhiFeatures.thirdSem = { presup := fun (x : E) => True, assertion := fun (x : E) => True }
Instances For
Person domain nesting: dom(1st) ⊆ dom(2nd) ⊆ dom(3rd).
Person nesting is a corollary of phiPresup_nesting — the same
theorem that derives number nesting also derives person nesting,
because both use the same PrivativePair structure.
Person and number have the same specLevel ordering — this is the
semantic content of @cite{harbour-2016}'s phi kernel isomorphism.
Both are phiPresup instances over the same PrivativePair cells,
so phiPresup_nesting applies to both: the nesting is structural,
not a per-domain coincidence.
§3b: Gender Presuppositions (@cite{sauerland-2003} §6) #
Gender features [±feminine, ±neuter] form a third PrivativePair instance,
with containment [+neuter] → [+feminine] (see Features.Gender.Features).
The presuppositional semantics mirrors number and person:
- neuter (maximal, specLevel 2): presupposes inanimate
- feminine (intermediate, specLevel 1): presupposes female
- masculine (minimal, specLevel 0): vacuous (default/unmarked)
@cite{wang-r-2023}: masculine, as the semantically unmarked gender, is available for honorific use cross-linguistically — paralleling the use of plural (unmarked number) and 3rd person (unmarked person) for politeness.
⟦Neut⟧: presupposes the referent is inanimate. Maximal cell [+feminine, +neuter] (specLevel 2).
Equations
- Semantics.Presupposition.PhiFeatures.neutSem isInanimate = { presup := isInanimate, assertion := fun (x : E) => True }
Instances For
⟦Fem⟧: presupposes the referent is female. Intermediate cell [+feminine, −neuter] (specLevel 1).
Equations
- Semantics.Presupposition.PhiFeatures.femSem isFemale = { presup := isFemale, assertion := fun (x : E) => True }
Instances For
⟦Masc⟧: vacuous presupposition. Minimal cell [−feminine, −neuter] (specLevel 0).
Equations
- Semantics.Presupposition.PhiFeatures.mascSem = { presup := fun (x : E) => True, assertion := fun (x : E) => True }
Instances For
Neuter features map to the maximal PrivativePair cell (specLevel 2).
Feminine features map to the intermediate cell (specLevel 1).
Masculine features map to the minimal cell (specLevel 0).
Gender domain nesting: dom(Neut) ⊆ dom(Fem) ⊆ dom(Masc). Parallels number (sg ⊆ pl) and person (1st ⊆ 3rd).
Gender nesting via phiPresup_nesting — structurally identical
to person nesting and number nesting.
Gender, person, and number have the same specLevel ordering —
all three domains share the phi kernel structure.
§4: Definiteness as Presupposition #
Definiteness exhibits the same presuppositional asymmetry as number and
person: definites carry a familiarity/uniqueness presupposition
(@cite{heim-1991}, @cite{strawson-1950}), while indefinites carry no
presupposition. Unlike number and person, definiteness is a binary
contrast (no intermediate cell), so we instantiate phiPresup at the
maximal and minimal cells only.
@cite{wang-r-2023} relies on this: indefinites are semantically unmarked (vacuous presupposition), so they are recruited for honorification in languages like Ainu.
⟦DEF⟧: presupposes the referent satisfies a contextual familiarity
or uniqueness condition. The predicate familiar is abstract —
concretely it may be Heim's familiarity or Russell's uniqueness
(cf. Features.Definiteness.DefPresupType).
Equations
- Semantics.Presupposition.PhiFeatures.defSem familiar = { presup := familiar, assertion := fun (x : E) => True }
Instances For
⟦INDEF⟧: no presupposition. Like plSem and thirdSem, its
distribution is constrained pragmatically by Maximize Presupposition.
Using an indefinite when a definite's presupposition is satisfied
would violate MP!.
Equations
- Semantics.Presupposition.PhiFeatures.indefSem = { presup := fun (x : E) => True, assertion := fun (x : E) => True }
Instances For
defSem is phiPresup at the maximal cell (with outerP = familiar).
Definiteness domain nesting: dom(DEF) ⊆ dom(INDEF).
The containment is strict: there exist unfamiliar entities in dom(INDEF) \ dom(DEF).
§5: Semantic Markedness (@cite{wang-r-2023}) #
A phi-feature value is semantically unmarked iff its presupposition is
vacuous — i.e., it is at the minimal PrivativePair cell (specLevel 0).
Semantically unmarked values are compatible with a wider range of
contexts, making them available for pragmatic co-optation (honorification).
This definition is domain-general: it applies uniformly to number (plural), person (3rd), and definiteness (indefinite).
A phi-feature value is semantically unmarked iff its specLevel is 0 (vacuous presupposition).
Equations
Instances For
A phi-feature value is semantically marked iff its specLevel is > 0 (substantive presupposition).
Equations
- Semantics.Presupposition.PhiFeatures.isSemanticMarked c = decide (c.specLevel > 0)
Instances For
The minimal cell is the unique unmarked cell.
The intermediate cell is marked.
Only the minimal cell is unmarked among well-formed cells.
Unmarked cells have vacuous presuppositions via phiPresup.
Well-formed cells have specLevel ≤ 2. This follows from the
three-cell structure of PrivativePair — the maximum is
maximal.specLevel = 2.
Presuppositional strength = specLevel. Higher specLevel = stronger presupposition = smaller domain.
Equations
Instances For
c₁ has a weaker presupposition than c₂.
Equations
- Semantics.Presupposition.PhiFeatures.presupWeakerThan c₁ c₂ = decide (c₁.specLevel < c₂.specLevel)
Instances For
c₁ has a stronger presupposition than c₂.
Equations
- Semantics.Presupposition.PhiFeatures.presupStrongerThan c₁ c₂ = decide (c₁.specLevel > c₂.specLevel)
Instances For
Minimal has the weakest presupposition among all cells.
Maximal has the strongest presupposition among all cells.
§7: Conceivability Lifting of Phi-Feature Presuppositions #
@cite{enguehard-2024} argues that number marking on indefinites triggers a conceivability presupposition: a singular indefinite presupposes it is conceivable the witness set has exactly one member; a plural indefinite presupposes it is conceivable the witness set has more than one member.
Conceivability is a general lifting on phi-feature presuppositions. For
any phiPresup cell with entity-level presupposition p, the
conceivability version asks whether p is satisfiable in a domain of
conceivable entities. This interacts with Maximize Presupposition: when
exactly one number's conceivability presupposition is satisfied, MP
forces that number; when both are satisfied, the choice is
underdetermined — opening the space for gradient probabilistic effects.
Entity-Level vs Cardinality Conceivability #
Two layers of conceivability arise:
Entity-level (
conceivableIn): some conceivable entity satisfies the presuppositionp. This is the general lifting applicable to all phi-feature presuppositions (number, person, definiteness).Cardinality (
cardConceivable): some conceivable situation has a witness set of the right cardinality. This is specific to indefinite number marking, where the presupposition is about the predicate's extension rather than a particular entity.
The two layers connect: if a conceivable situation has exactly one witness, that witness is atomic; if ≥ 2, their sum is non-atomic.
Connection to PresuppositionContext.presupSatisfiable #
Conceivability = satisfiability in context:
presupSatisfiable c p from Core.Semantics.PresuppositionContext
checks whether p.presup is compatible with context set c. The
conceivability presupposition of an indefinite is the meta-requirement
that the number feature's entity-level presupposition be satisfiable.
General conceivability: a property p is conceivable over a domain
C iff some entity in C satisfies it. This lifts entity-level
presuppositions to context-level satisfiability requirements.
Equations
- Semantics.Presupposition.PhiFeatures.conceivableIn p C = ∃ (e : E), C e ∧ p e
Instances For
Conceivability lifting of a phi-feature presupposition: it is
conceivable that the presupposition of cell c could be satisfied
by some entity in domain C.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The minimal cell is always conceivable over any non-empty domain (vacuous presupposition).
Conceivability nests: if a more specified cell is conceivable,
all less specified cells are too. Mirrors phiPresup_nesting.
Sg conceivability ↔ some conceivable entity is atomic.
Actuality implies conceivability: if p e holds and e ∈ C, then
p is conceivable in C. Standard presuppositions are special cases
of conceivability presuppositions where the actual referent is in
the conceivable domain.
Conceivability is monotone in the domain: enlarging the set of conceivable entities preserves conceivability.
Witness-cardinality conceivability: some conceivable situation has a
witness set whose cardinality satisfies cardPred.
W is a type of conceivable situations, witnessCard gives the
cardinality of the indefinite's witness set in each situation, and
cardPred is the number feature's cardinality requirement.
This generalizes the binary sg/pl contrast: for dual,
cardPred = (· = 2); for paucal, cardPred = (2 ≤ · ∧ · ≤ 5).
Equations
- Semantics.Presupposition.PhiFeatures.cardConceivable witnessCard conceivable cardPred = ∃ (w : W), conceivable w ∧ cardPred (witnessCard w)
Instances For
Sg indefinite conceivability: ∃ conceivable situation with exactly one witness (@cite{enguehard-2024} generalization 7).
Equations
- Semantics.Presupposition.PhiFeatures.sgCardConceivable witnessCard conceivable = Semantics.Presupposition.PhiFeatures.cardConceivable witnessCard conceivable fun (x : ℕ) => x = 1
Instances For
Pl indefinite conceivability: ∃ conceivable situation with ≥ 2 witnesses (@cite{enguehard-2024} generalization 7).
Equations
- Semantics.Presupposition.PhiFeatures.plCardConceivable witnessCard conceivable = Semantics.Presupposition.PhiFeatures.cardConceivable witnessCard conceivable fun (x : ℕ) => x ≥ 2
Instances For
Du indefinite conceivability: ∃ conceivable situation with exactly 2 witnesses.
Equations
- Semantics.Presupposition.PhiFeatures.duCardConceivable witnessCard conceivable = Semantics.Presupposition.PhiFeatures.cardConceivable witnessCard conceivable fun (x : ℕ) => x = 2
Instances For
Sg and pl conceivability are not complementary: both can hold when the conceivable domain contains situations of both kinds. This is the structural reason MP is underdetermined in intermediate cases.
When ALL conceivable situations have unique witnesses, only sg is conceivable — pl conceivability fails.
When ALL conceivable situations have multiple witnesses, only pl is conceivable — sg conceivability fails.
Conceivability presuppositions are incomparable: neither sg's entails pl's nor vice versa. This means standard Maximize Presupposition (which requires a strength ordering) cannot select between them — a structural gap that gradient probabilistic effects fill (@cite{enguehard-2024} §4.1).
In negative contexts (witness set empty in the actual situation), the conceivability presupposition is non-trivially about OTHER conceivable situations — the actual one witnesses neither sg nor pl conceivability. This is why conceivability presuppositions are observable under negation while scalar implicatures are not.
Du conceivability entails pl conceivability: if exactly 2 is conceivable, then ≥ 2 is conceivable.
Conceivability is monotone in the conceivable domain: enlarging the set of conceivable situations preserves conceivability.