@cite{percus-2000}: Constraints on Situation Variables in Syntax @cite{percus-2000} #
@cite{heim-kratzer-1998} @cite{kratzer-1998} @cite{partee-1973}
Formalizes Percus's theory of situation pronouns in LF. Every predicate takes a situation argument, every clause introduces a lambda-s binder, and Generalization X constrains which binder can bind which variable.
Generalization X #
The situation pronoun that a predicate is associated with must be bound by the minimal c-commanding situation binder.
This is a syntactic well-formedness constraint that predicts:
- Main predicates in attitude complements are obligatorily de dicto (evaluated in belief situations, not the actual situation)
- NP restrictors can be de re (their situation variable can be bound by a higher lambda-s)
- "Mixed" readings where the main predicate is de re but the NP is de dicto are impossible
Tower Bridge #
Under the ContextTower analysis, Generalization X becomes a depth constraint:
the main predicate of an embedded clause must access the situation at
DepthSpec.local (the most deeply embedded layer), while NP restrictors
are unconstrained (they may access any depth). This connects Percus's
syntactic constraint to the depth-indexed access pattern system.
Situation Assignment Infrastructure #
Situation assignments specialize Core.VarAssignment from D = Time
(Partee's temporal variables) to D = Situation W Time (Percus's
situation variables). The algebraic structure is identical:
| @cite{partee-1973} | @cite{percus-2000} |
|---|---|
TemporalAssignment | SituationAssignment |
interpTense n g | interpSitVar n g |
updateTemporal g n t | updateSitVar g n s |
temporalLambdaAbs | sitLambdaAbs |
Situation assignment function: maps variable indices to situations.
The situation-level analogue of Partee's TemporalAssignment
and H&K's entity assignment Assignment.
Equations
Instances For
Situation variable denotation: s_n^g = g(n).
Specializes Core.VarAssignment.lookupVar.
Equations
Instances For
Modified situation assignment g[n -> s].
Specializes Core.VarAssignment.updateVar.
Equations
Instances For
Situation lambda abstraction: bind a situation variable.
Specializes Core.VarAssignment.varLambdaAbs.
Every clause boundary introduces a lambda-s_n binder in Percus's system. Under attitude verbs, the binder's value is filled by quantification over doxastic alternatives.
Equations
Instances For
A predicate binding record: which situation variable a predicate uses and which binder (by index) should bind it.
In Percus's LF, a sentence like: [lambda-s1 Mary believes_s1 [lambda-s2 John isCanadian_s2]] has the predicate "isCanadian" bound by lambda-s2 (index 2). Generalization X says this binding is the ONLY well-formed option: the predicate must use the closest c-commanding binder.
- sitVarIndex : ℕ
The situation variable index the predicate uses
- closestBinderIndex : ℕ
The index of the closest c-commanding lambda-s binder
Instances For
Generalization X:
The situation pronoun that a verb/predicate is associated with must be bound by the minimal c-commanding situation binder.
A predicate's binding is Gen-X-compliant iff its situation variable is bound by the closest binder.
Equations
- b.genXCompliant = (b.sitVarIndex == b.closestBinderIndex)
Instances For
An LF reading is well-formed under Generalization X iff ALL predicate bindings use their closest c-commanding binder.
Equations
Instances For
Generalization Y (@cite{percus-2000}, p. 204, example 39):
The situation pronoun that an adverbial quantifier selects for must be coindexed with the nearest lambda above it.
Gen Y is a parallel constraint to Gen X, but for adverbial quantifiers ("always", "usually", "never") rather than predicates (verbs, adjectives).
The situation pronoun ssh that a quantifier like "always" uses to determine its domain must be bound by the nearest c-commanding lambda-s. This prevents "always" from reaching past an attitude verb to quantify over actual-world situations rather than belief-world situations.
The PredicateBinding structure is reused: it records which
situation variable the quantifier uses and which binder is closest.
Equations
- Semantics.Intensional.Situations.Percus.genYWellFormed quantifierBindings = quantifierBindings.all Semantics.Intensional.Situations.Percus.PredicateBinding.genXCompliant
Instances For
Combined well-formedness: both Gen X (predicates) and Gen Y (adverbial quantifiers) must use their closest binder.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Tower formulation of Generalization X: a situation access pattern
for a main predicate must read from .local (the most deeply
embedded context, corresponding to the closest lambda-s binder).
NP restrictors are unconstrained — they may read from any depth,
including .origin (de re) or .relative k (intermediate).
Equations
Instances For
NP restrictors can access any depth (no constraint).
Instances For
A tower-based LF reading is Gen-X-compliant iff all predicate access patterns read from local and restrictors are unconstrained.
Equations
- Semantics.Intensional.Situations.Percus.genXTowerWellFormed predicatePatterns _restrictorPatterns = ∀ p ∈ predicatePatterns, Semantics.Intensional.Situations.Percus.GenXAsTowerDepth p.snd
Instances For
Bridge: a PredicateBinding where sitVarIndex == closestBinderIndex corresponds to GenXAsTowerDepth (depth =.local). Both express the same constraint: the predicate reads from the nearest binder.
In PredicateBinding, the nearest binder is identified by index equality.
In the tower, the nearest binder is the innermost context (.local).
These are the same notion, expressed in different frameworks.
Doxastic alternatives as situations.
dox agent s returns the situations compatible with what agent
believes at situation s. Generalizes Hintikka's Dox_x(w) from
worlds to world-time pairs.
Equations
- Semantics.Intensional.Situations.Percus.DoxSit W Time E = (E → Situation W Time → List (Situation W Time))
Instances For
Attitude verb semantics with situation binding.
s_n^g(s) = forall s' in Dox_x(s). phi^(g[n -> s'])
The attitude verb quantifies universally over doxastic alternatives. Each alternative s' is bound to situation variable n in the complement. Predicates inside the complement that reference s_n are thereby evaluated in belief situations — this is the de dicto reading.
Predicates that reference a DIFFERENT variable (e.g., s_m where m != n) escape the binding and are evaluated at whatever s_m is set to by the outer binder — this would be a de re reading. Generalization X blocks this for main predicates but allows it for NP restrictors.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Adverbial quantifier "always" with situation binding.
always_ssh [lambda-s_n. phi]^g = forall s' in domain(ssh). phi^(g[n -> s'])
The quantifier introduces a binder lambda-s_n over its nuclear scope. Its domain is determined by the situation pronoun ssh: the set of relevant situations (game rounds, time points, etc.) at the world of ssh.
Generalization Y constrains ssh: it must be bound by the nearest c-commanding lambda — typically the one introduced by an attitude verb when "always" is embedded under an attitude.
Equations
- Semantics.Intensional.Situations.Percus.alwaysAt domain ssh n scope g = (domain ssh).all fun (s' : Situation W Time) => scope (Semantics.Intensional.Situations.Percus.updateSitVar g n s')
Instances For
Bound situation variable receives the binder's value.
Parallel to zeroTense_receives_binder_time in Core/Tense.lean.
Non-target variables are unaffected by binding.
Project a situation assignment to a temporal assignment.
This is Core.Tense.situationToTemporal with a transparent
definition: we extract the temporal coordinate from each situation
(@cite{heim-kratzer-1998}, formula 38: time(e) on eventualities/situations).
Equations
Instances For
Situation assignment projects to temporal assignment coherently: the temporal variable at index n equals the time of the situation at index n.