TL;DR
This paper explores how the meaning of quantifiers like "always" and "usually" is heavily influenced by context, much like pronouns. It argues that the way we restrict the domain of these quantifiers relies on similar "accessibility" mechanisms as anaphora (pronoun resolution), suggesting a deeper, shared grammatical and semantic architecture for handling context-dependent meaning.
What Problem Does This Paper Address?
Ever wonder how we understand sentences like "He usually travels by car"? The word "usually" doesn't apply to all of his travels, but to a specific subset. The problem this paper tackles is understanding how this subset, or "domain," is determined. It's not explicitly stated in the sentence; instead, it's implicitly narrowed down by the surrounding discourse and our general knowledge. This phenomenon, known as "domain restriction" for quantifiers, is complex because it's highly context-dependent. The paper investigates the underlying linguistic mechanisms that allow us to resolve these implicit restrictions, drawing parallels to how we resolve pronouns (anaphora), which also rely on identifying the correct antecedent from a pool of possibilities. The challenge lies in characterizing the interface between sentence grammar and the broader discourse context that shapes these interpretations.
Methods
The paper doesn't propose new computational algorithms or statistical models. Instead, it uses a linguistic analysis approach. It examines examples of natural language discourse to illustrate how quantificational domains are restricted. The core method involves:
- Comparative Analysis: Comparing different discourse structures to show how subtle changes affect the interpretation of quantifiers. For instance, the order of sentences in examples (2a) and (2b) drastically changes the perceived domain of "usually."
- Drawing Analogies: Explicitly linking the mechanisms of domain restriction for quantifiers to established principles of anaphora resolution (e.g., "precede/command" conditions) and presupposition projection.
- Identifying "Accessibility" Constraints: Proposing that a common notion of "accessibility" governs what information in the discourse or context is available to resolve both quantificational domains and anaphoric references.
Datasets
Not discussed in the paper. The paper relies on constructed linguistic examples rather than formal datasets.
Key Insights and Ideas
- Quantifier Domains are Dynamic: The meaning of quantifiers like "usually" isn't fixed but is dynamically shaped by the surrounding discourse.
- Anaphora and Quantifiers Share Mechanisms: The way we resolve pronouns (anaphora) and restrict the domains of quantifiers relies on similar underlying "accessibility" constraints. This suggests a unified system for handling context-dependent meaning.
- "Backwards Domain Restriction" is Constrained: Just as pronouns can't always refer "backwards" to antecedents that appear later in a sentence, the domains of quantifiers also face similar constraints when trying to restrict themselves to information presented later in the discourse.
- Local Contexts Matter: Beyond the overall discourse, "local contexts" within sentences (influenced by syntax and semantics) also play a role in shaping quantifier interpretation.
- Interface Between Grammar and Context: A key issue is understanding the precise nature of the interaction between grammatical rules and extragrammatical (discourse, world knowledge) mechanisms in resolving these ambiguities.
Key Results
The paper presents qualitative findings based on linguistic examples. The "results" are the insights derived from analyzing these examples:
- Discourse structure significantly impacts the interpretation of quantificational domains. For example, in (2b), the discourse is perceived as contradictory because the domain of "usually" cannot be restricted to the specific trips mentioned later.
- The constraints on "backwards domain restriction" for quantifiers are analogous to those found in anaphora.
- The paper highlights that while semantic and syntactic structures often align, the underlying accessibility constraints are fundamentally semantic in nature, as demonstrated by work on presupposition projection.
Limitations and Future Work
- Formalization: The paper focuses on articulating relationships and identifying phenomena rather than proposing a specific formal model. The authors acknowledge that the "interface between the grammatical and the extragrammatical mechanisms" is a complex area needing further characterization.
- Future Directions: The paper points to ongoing work by Rooth and Von Fintel on relating focus structure to anaphoric structure, and the broader field of dynamic semantics as promising avenues for future research. It suggests that a deeper understanding of these "recursive contexts" and their shared accessibility constraints is crucial.
Summary updated