Kratzer & Shimoyama (2002): Indeterminate Pronouns #
@cite{kratzer-shimoyama-2002}
"Indeterminate Pronouns: The View from Japanese." In C. Lee et al. (eds.), Contrastiveness in Information Structure, Alternatives and Scalar Implicatures, Studies in Natural Language and Linguistic Theory 91, 123-143.
Core Thesis #
Hamblin alternative semantics, originally designed for questions, is the right architecture for all quantification. Indeterminate pronouns (Japanese dare, nani, doko...) denote sets of individual alternatives that expand pointwise via functional application until caught by an operator (∃, ∀, Neg, Q). Quantification is operator selection, not DP movement.
Formalized Contributions #
- Hamblin operators (§2): The four sentential operators over propositional alternative sets.
- Pointwise FA = Set applicative (§3): K&S's Hamblin FA is exactly
the set applicative from @cite{charlow-2020}, already formalized in
Applicative.lean. - GQ as special case (§2): Determiner quantification falls out
when alternatives are individuals —
[∃]({P(x) : x ∈ A}) ↔ ∃x∈A, P(x). - Singleton collapse: When alternatives are a singleton (ordinary semantics), Hamblin modals reduce to standard Kripke modals.
- Modal–indefinite interaction (§7): Possibility/necessity modals are sensitive to Hamblin alternatives in their scope.
- Distribution requirement as implicature (§6, §8): The free choice effect is derived via Gricean reasoning, not semantic entailment.
- End-to-end FC derivation: Hamblin T-content + implicature = FC.
- Selectivity (§9): Non-selective (Japanese) vs. selective (Indo-European) indeterminate systems, with Beck effect data.
- Cross-linguistic paradigm (§1): Latvian indeterminate series.
Integration Points #
- §3 Hamblin FA bridges to
setAp(Composition/Applicative.lean) - Singleton collapse bridges Hamblin modals to Kripke semantics
- §8 free choice bridges to
free_choice_forward(Exhaustification/FreeChoice.lean) - Fragment data bridges to
Japanese/Determiners.lean,German/ModalIndefinites.lean, andLatvian/IndeterminatePronouns.lean
§2-3: Hamblin Interpretation System #
In a Hamblin semantics, all expressions denote sets of alternatives. Most lexical items denote singleton sets; indeterminate pronouns denote sets of individuals. Composition is pointwise functional application.
A Hamblin denotation is a set of alternatives of type α. This is exactly the carrier of @cite{charlow-2020}'s set monad.
Equations
Instances For
Hamblin Functional Application (§3): pointwise application of a set of functions to a set of arguments.
⟦α⟧ = {a ∈ Dσ : ∃b ∈ ⟦β⟧ ∃c ∈ ⟦γ⟧, a = c(b)}
Equations
- Phenomena.ModalIndefinites.Studies.KratzerShimoyama2002.hamblinFA funSet argSet b = ∃ (f : A → B), funSet f ∧ ∃ (x : A), argSet x ∧ f x = b
Instances For
Bridge: Hamblin FA = the set applicative from Applicative.lean.
§2: Sentential Operators over Propositional Alternatives #
The alternatives created by indeterminate phrases expand until caught by an operator. Where A is a set of propositions (p. 126-127):
[∃](A)— true iff some proposition in A is true[∀](A)— true iff every proposition in A is true[Neg](A)— true iff no proposition in A is true[Q](A)— A itself (the Hamblin question denotation)
[∃](A): existential closure over propositional alternatives.
Equations
- Phenomena.ModalIndefinites.Studies.KratzerShimoyama2002.opExists A w = ∃ (p : W → Prop), A p ∧ p w
Instances For
[∀](A): universal closure over propositional alternatives.
Equations
- Phenomena.ModalIndefinites.Studies.KratzerShimoyama2002.opForall A w = ∀ (p : W → Prop), A p → p w
Instances For
[Neg](A): negative closure over propositional alternatives.
Equations
- Phenomena.ModalIndefinites.Studies.KratzerShimoyama2002.opNeg A w = ∀ (p : W → Prop), A p → ¬p w
Instances For
[Q](A): question operator — identity on propositional alternatives.
Instances For
Neg is the pointwise negation of ∃: [Neg](A)(w) ↔ ¬[∃](A)(w).
∀ entails ∃ on non-empty alternative sets.
Map from operator tags to their semantic implementations.
- exists_ : QuantOperator
- forall_ : QuantOperator
- neg : QuantOperator
- question : QuantOperator
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Semantic interpretation of a propositional quantificational operator.
Returns none for .question, which produces an alternative set rather
than a proposition.
Equations
- Phenomena.ModalIndefinites.Studies.KratzerShimoyama2002.QuantOperator.exists_.applyProp A = some (Phenomena.ModalIndefinites.Studies.KratzerShimoyama2002.opExists A)
- Phenomena.ModalIndefinites.Studies.KratzerShimoyama2002.QuantOperator.forall_.applyProp A = some (Phenomena.ModalIndefinites.Studies.KratzerShimoyama2002.opForall A)
- Phenomena.ModalIndefinites.Studies.KratzerShimoyama2002.QuantOperator.neg.applyProp A = some (Phenomena.ModalIndefinites.Studies.KratzerShimoyama2002.opNeg A)
- Phenomena.ModalIndefinites.Studies.KratzerShimoyama2002.QuantOperator.question.applyProp A = none
Instances For
The question operator returns the alternative set unchanged:
[Q](A) = A. This is distinct from the propositional operators
because it does not collapse alternatives to a truth value.
Determiner quantification as special case (p. 126): "Determiner quantification falls out as a special case, the case where the alternatives are individuals."
When an indeterminate denotes a set of individuals A and a predicate
lifts each individual to a proposition, sentential [∃] over the
resulting propositional alternatives equals the standard GQ existential:
[∃]({P(x) : x ∈ A})(w) ↔ ∃ x ∈ A, P(x)(w).
Universal counterpart: [∀] over individual alternatives = standard ∀.
[∀]({P(x) : x ∈ A})(w) ↔ ∀ x ∈ A, P(x)(w).
Compositional Derivation of dare(-ga) nemutta #
Japanese indeterminate pronouns denote sets of individuals. Composed with a predicate via pointwise FA, they produce propositional alternative sets. An operator then closes the set (p. 126):
⟦dare⟧^{w,g} = {x: human(x)(w)}⟦nemutta⟧^{w,g} = {λxλw'. slept(x)(w')}(singleton)⟦dare nemutta⟧^{w,g} = {p: ∃x[human(x)(w) & p = λw'. slept(x)(w')]}
We simplify by working extensionally (dropping the world parameter on the restrictor), which is faithful for the core point that operator selection = quantification.
Equations
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
⟦dare⟧ = the set of all humans (extensional simplification).
Instances For
⟦nemutta⟧ = singleton set containing the sleep predicate.
Equations
Instances For
⟦dare nemutta⟧ = {slept(a), slept(b), slept(c)} via Hamblin FA.
Equations
- One or more equations did not get rendered due to their size.
Instances For
dare-ka nemutta = [∃]({slept(a), slept(b), slept(c)}) = ∃x.slept(x)
dare-mo nemutta = [∀]({slept(a), slept(b), slept(c)}) = ∀x.slept(x)
§7: Modals over Hamblin Alternative Sets #
The key insight: modals can be sensitive to the propositional alternatives introduced by indeterminate phrases in their scope (p. 132-133).
Possibility/necessity modals over an alternative set A:
⟦kann α⟧(w) = ∃w'[R(w,w') ∧ ∃p[p ∈ A ∧ p(w')]]
⟦muss α⟧(w) = ∀w'[R(w,w') → ∃p[p ∈ A ∧ p(w')]]
The distribution requirement (to be derived as implicature in §8):
∀p[p ∈ A → ∃w'[R(w,w') ∧ p(w')]]
distributes alternatives over accessible worlds.
Note: We use Prop-valued accessibility here (rather than the Bool-valued
Core.ModalLogic.AccessRel) to stay in Prop throughout the Hamblin
semantics. The singleton collapse theorem below shows these Hamblin modals
reduce to standard Kripke modals when the alternative set is a singleton.
Prop-valued accessibility relation for Hamblin modal semantics.
Named distinctly from Core.ModalLogic.AccessRel (which is Bool-valued)
to avoid shadowing.
Equations
Instances For
Possibility modal over Hamblin alternatives (§7, p. 133): True at w iff some accessible world satisfies some alternative.
Equations
- Phenomena.ModalIndefinites.Studies.KratzerShimoyama2002.hamblinPoss R A w = ∃ (w' : W), R w w' ∧ ∃ (p : W → Prop), A p ∧ p w'
Instances For
Necessity modal over Hamblin alternatives (§7, p. 133): True at w iff every accessible world satisfies some alternative.
Equations
- Phenomena.ModalIndefinites.Studies.KratzerShimoyama2002.hamblinNec R A w = ∀ (w' : W), R w w' → ∃ (p : W → Prop), A p ∧ p w'
Instances For
The distribution requirement (§7, p. 133): for every alternative p in A, there exists an accessible world where p is true.
Equations
- Phenomena.ModalIndefinites.Studies.KratzerShimoyama2002.distribReq R A w = ∀ (p : W → Prop), A p → ∃ (w' : W), R w w' ∧ p w'
Instances For
Singleton collapse: when the alternative set is a singleton {p}, Hamblin possibility reduces to standard Kripke possibility. This is the paper's core architectural claim: ordinary semantics is the special case where all denotations are singletons.
Singleton collapse for necessity: when alternatives are a singleton, Hamblin necessity reduces to standard Kripke necessity.
Necessity entails possibility (when some accessible world exists).
The distribution requirement is NOT entailed by necessity (§6, p. 131). Necessity only requires some alternative per world, not that every alternative is witnessed. The distribution requirement is an implicature.
Countermodel: R reflexive-only, A = {p₁, p₂} where p₁ holds at true, p₂ holds at false. From w = true, only true is accessible: necessity holds (p₁ witnesses true) but distribution fails (p₂ is unwitnessed).
§7: Domain Widening #
ein Mann denotes a contextually restricted subset of men (Schwarzschild 2000: singleton indefinites). irgendein Mann widens to the full set (p. 132).
This is the same mechanism as contextual domain restriction in
DomainRestriction.lean: ein selects from a contextually restricted
domain C ∩ P, while irgend- removes the restriction.
A simple indefinite selects from a contextually restricted subset.
Equations
Instances For
Widening weakens existentials: restricted entails widened.
§6 & §8: Pragmatic Derivation of the Free Choice Implicature #
§6 establishes that the distribution requirement is a conversational implicature: cancelable (ex. 11), disappears in DE contexts (ex. 12, 14).
§8 derives it via Gricean reasoning about why the speaker widened. Widening could serve: (a) strengthening, (b) avoiding a false claim, (c) avoiding a false exhaustivity inference (p. 134).
Three cases over alternatives {A, B}:
(16) Possibility: Du kannst dir irgendeins leihen #
- T-content: ◇(A ∨ B)
- Implicature: ◇A ↔ ◇B
- Total: ◇A ∧ ◇B
(17) Necessity: Du musst dir irgendeins leihen #
- T-content: □(A ∨ B)
- Implicature: □A ↔ □B
- Total: □(A ∨ B) ∧ ◇A ∧ ◇B
(18) Negated possibility: auf keinen Fall #
- T-content: ¬◇(A ∨ B)
- No implicature: canceled (widening adds nothing in DE context)
(17) Necessity total meaning (p. 135). □(A∨B) → ◇(A∨B) → ◇A ∨ ◇B, combined with ◇A↔◇B, gives □(A∨B) ∧ ◇A ∧ ◇B.
End-to-end FC derivation for (16): Given two propositional
alternatives under a possibility modal, the T-content is exactly
hamblinPoss, and applying the biconditional implicature yields FC.
This connects the modal semantics (§7) to the pragmatic derivation (§8) in a single theorem.
Bridge to @cite{chierchia-2013}. K&S's pragmatic derivation (Gricean reasoning) and Chierchia's grammatical derivation (double exhaustification) both yield ◇A ∧ ◇B. Different mechanisms, same empirical prediction.
§9: Non-Selective vs. Selective Indeterminate Systems #
Japanese: non-selective — same base (dare) + different particles (ka → ∃, mo → ∀, demo → FC). Base does not change shape.
Indo-European: selective — irgendein associates only with [∃], not [∀], [Neg], or [Q]. Explained via uninterpretable features (p. 138): selective indeterminates carry uninterpretable [∃] that must be checked against an interpretable counterpart via feature movement.
Beck Effects (§9, p. 139) #
When feature movement of uninterpretable [∃] is blocked by an intervening scope-bearing element, ungrammaticality results:
- (23a) *Was hat sie nicht WEM gezeigt? — blocked by nicht ([Neg])
- (23b) *Was hat sie nie WEM gezeigt? — blocked by nie
- (23c) *Was hat niemand WEM gezeigt? — blocked by niemand
- (23d) *Was hat fast jeder WEM gezeigt? — blocked by fast jeder
- (23e) *Was hat (irgend)jemand WEM gezeigt? — blocked by jemand
- (23f) Was hat der Hans WEM gezeigt? — OK (definite: no scope feature)
- (23g) Was hat sie damals WEM gezeigt? — OK (adverb: no scope feature)
An indeterminate pronoun paradigm: which operators it associates with, and whether its morphology changes per operator.
- language : String
- base : String
- associatesWith : List QuantOperator
- morphologicallyMarked : Bool
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- p.isNonSelective = decide (p.associatesWith.length ≥ 3)
Instances For
Equations
- p.isSelective = decide (p.associatesWith.length ≤ 2)
Instances For
Japanese dare: non-selective. Associates with all four operators via different particles. Base form does not change.
Equations
- One or more equations did not get rendered due to their size.
Instances For
German irgend-: selective. Associates only with [∃] (§9, p. 137). Cannot associate with [∀] (ex. 20c), [Neg] (ex. 21), or [Q].
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
Equations
- One or more equations did not get rendered due to their size.
- Phenomena.ModalIndefinites.Studies.KratzerShimoyama2002.instBEqInterventionDatum.beq x✝¹ x✝ = false
Instances For
Beck effect paradigm (examples 23a-g): scope-bearing elements block feature movement of [∃]/[Q]; non-scope-bearing elements don't.
Pattern: *Was hat sie [INTERVENER] WEM gezeigt?
Equations
- One or more equations did not get rendered due to their size.
Instances For
Scope-bearing elements block; non-scope-bearing elements don't.
The generalization: scope-bearing = ungrammatical, non-scope-bearing = OK.
Same base (dare), different force via particle alternation.
dare-ka is existential; paradigm predicts ∃ association.
dare-mo is universal; paradigm predicts ∀ association.
irgendein is existential-only + not-at-issue (domain widening).
§1: Indeterminate Pronoun Paradigms Cross-Linguistically #
@cite{haspelmath-1997} (p. 277, diacritics omitted). The Latvian paradigm illustrates a selective system: each operator association is morphologically marked by a distinct prefix (kaut- existential, ne- under direct negation, jeb- indirect negation/comparatives/FC).
Latvian paradigm data imported from Fragments/Latvian/IndeterminatePronouns.lean.
Latvian is morphologically marked (selective); Japanese is not.