Fact 1: Quantifier scope in natural language is constrained
A. Structural constraints ('scope islands')
B. Scope freezing constructions
C. Semantic constraints
Fact 2: Indefinites do not abide by these constraints
The most famous case is the English double object construction (Larson 1990)
Compare
Inverse scope readings are significantly harder when downward entailing quantifiers are involved (cf. Mayr & Spector 2012)
Standard assumption: the same scope shifting mechanism (e.g. Quantifier Raising, type-shifting) applies to all quantificational noun phrases
Issue: If we assume that the scope shifting mechanism is constrained (structurally and semantically), then we undergenerate for indefinites with exceptional wide scope
Standard answer: A different mechanism for indefinites' exceptional wide scope, e.g.
Non-standard answer: Different constraints for different scopal elements (Barker 2022)
Definite noun phrases pragmatically presuppose uniqueness (unique maximality, more generally), sometimes with respect to some discourse information (recall
Partitive indefinites (e.g. some of my supervisees) have definite domains; the scope of the indefinite part is independent
'Epistemic specificity' (Farkas 1995, 2002a,b), e.g. a certain, "ある", koe-wh indefinites in Russian
'Specific indefinite' has never been given a clear definition (Farkas 1995, 2002a,b)
Differential Object Marking in languages like Turkish forces wide scope
Ali
Ali
bir
one
piyano(-yu)
piano(-
kiralamak
rent.
istiyor.
wants
In addition, definite objects are obligatorily accusative marked
Definite and indefinite noun phrases assert existence; definite noun phrases presupposes uniquenes (existence + 'at most one')
'Specific indefinites' presuppose existence
definite | specific indefinite | plain indefinite | |
---|---|---|---|
presupposition | unique | existence | none |
English articles | the | a/some | |
Turkish objects | X | accusative | no case (nominative) |
Japanese nouns | X | bare nouns |
A context is a set of total assignments
Sentences denote Context Change Potentials (CCPs) = functions over contexts
A (plain) indefinite statement is eliminative
We can assume that all noun phrases are associated with a new variable
A plain indefinite has no presupposition
A definite statement has a uniqueness presupposition (to be revised)
This is a unique definite
Notation:
Familiar definite
Bridging definites are analyzed in an analogous way with a free variable, but let's omit them to avoid complications that arrise with free relations R
Indefinite and definite noun phrases assert the same thing (as in Heim 1982)
Both definite and indefinite noun phrases are associated with new variables
These new variables are eliminable by moving to a stack-based setup, instead of assignments (cf. Van Eijck 2001, Nouwen 2003, 2007)
Familiar and bridging definites involve old variables (naturally!)
For Heim 1982, definites are associated with old variables and do not trigger uniqueness presuppositions (but familiarity presuppositions)
Proposal: Specific indefinites presuppose existence
To understand what this means, recall first that anaphoric meaning cannot be reduced to propositional meaning (Karttunen 1976, Heim 1982, Sudo 2023)
To capture this, Heim's 1982 File Change Semantics operates on two types of model theoretical objects: possible worlds and assingment functions
For Heim 1982, presuppositions are static propositions that need to be commonly known to be true, i.e., they require all possible worlds in
I propose in addition that presuppostions may carry new anaphoric information (dynamic presuppositions) (Beaver 1992, Elliott & Sudo 2021, Mayr & Sudo 2020)
Notation:
Heimian presupposition
Dynamic presupposition
This change is propositionally inert, but crucially enables dynamic binding from
Cf. Stalnaker's pragmatics of assertion and presuppositions
Notation:
Plain indefinite
Specific indefinite
The pragmatic presupposition is simply existential and propositional
In a simple positive sentence like this, there is no effect in the assertion
Negation
Explaining (selective) quantifiers will take a lot of time, so let's talk about projection through quantifiers schematically
Fodor & Sag 1982 claimed that wide scope indefinites always take maximal scope, and there is no intermediate scope readings, based on examples like:
But other examples have intermediate scope readings.
Note the bound pronouns!!
It has been remarked that bound pronouns 'facilitate' intermediate scope readings
Presuppositions project through quantifiers universally, cf. Every boy quit smoking
The existence presupposition of (1a): Every teacher has a
The assertion of (1a): c[Every teacher has a
For some cases, it might not be too far-fetched to postulate a covert relational variable with an old index (e.g., "R y") but for all
Furthermore:
There's no way to analyse this in terms of quantificational subordination
One possible solution is (intermediate) presupposition accommodation
More research needed to see if the availability of intermediate scope readings correlates with the availability of (intermediate) presupposition accommodation, as predicted by this analysis
A neo-Heimian dynamic semantics for (in)definiteness
All indefinite and definite noun phrases are associated with new variables (= they function like existential quantifiers)
They differences are in the presuppositions
Dynamic presuppositions allow for a new type of indefinites with existence presuppositions ('specific indefinites')
definite | specific indefinite | plain indefinite | |
---|---|---|---|
presupposition | unique | existence | none |
English articles | the | a/some | |
Turkish objects | X | accusative | no case (nominative) |
Japanese nouns | X | bare nouns |