Dynamic semantics

Dynamic semantics is a family of formal theories in linguistic semantics that represents meaning through the capacity of an expression to transform an informational context. In a static semantics, a sentence ordinarily denotes a truth value relative to a model and an assignment. In a dynamic semantics, the same sentence denotes an operation or relation connecting an input context with the contexts compatible with the information it contributes.

The dynamic approach was developed to explain dependencies that extend across sentence boundaries, particularly those involving anaphora, indefinites, and presupposed information. Its central formal insight is that interpretation modifies the structures against which subsequent expressions are evaluated. The resulting account integrates truth conditions with the management of discourse information without identifying linguistic meaning with either conversational intention or psychological processing.

Context change

A context may be represented as a set of possible worlds, a set of variable assignments, or a more articulated information state containing both kinds of object. If (c) is a context and (\varphi) is a sentence, the update of (c) with (\varphi) is conventionally written as

[ c[\varphi]. ]

For a nonmodal declarative sentence, the update commonly removes those possibilities in which the sentence is false. When contexts are sets of possible worlds, a basic update rule has the form

[ c[\varphi] = {w \in c \mid \varphi \text{ is true in } w}. ]

This operation resembles intersection with the proposition expressed by (\varphi), but the dynamic framework becomes substantively different from an ordinary propositional account when contexts contain discourse referents or assignments. An indefinite expression can then extend the context with a value that remains available to a later pronoun.

Consider the discourse:

A researcher entered. She sat down.

The first sentence introduces a discourse referent associated with a researcher who entered. The second sentence tests whether the same referent sat down. A compositional analysis can therefore preserve the dependency between the indefinite noun phrase and the pronoun even though the two expressions occur in separate sentences.

Dynamic conjunction is interpreted through sequential update:

[ c[\varphi \land \psi] = \bigl(c[\varphi]\bigr)[\psi]. ]

The intermediate state produced by (\varphi) serves as the input for (\psi). Consequently, dynamic conjunction need not behave as a commutative operation at the level of discourse interpretation, even when the completed discourse has classical truth conditions. In “A researcher entered and she sat down,” the indefinite becomes available before the pronoun is interpreted. Reversing the clauses does not generally preserve that accessibility relation.

Historical development

The conceptual background of dynamic semantics includes Robert Stalnaker’s account of assertion as the reduction of a conversational context set. Lauri Karttunen subsequently analyzed discourse referents and the conditions under which they remain available for anaphoric reference. These developments established the idea that semantic interpretation must track changes in discourse information rather than evaluate every sentence in isolation.

Hans Kamp introduced discourse representation theory as a structured account of discourse referents, accessibility, and truth conditions. A discourse is translated into a discourse representation structure whose referents and conditions encode the information accumulated during interpretation. Accessibility constraints determine whether a later expression can retrieve a referent introduced inside negation, implication, or another subordinate construction.

In a separate formulation, Irene Heim developed file change semantics, in which indefinites update a file of assignment possibilities and pronouns retrieve values already represented in that file. Heim’s treatment connected anaphoric accessibility with the semantic behavior of existential quantification. It also supplied a dynamic analysis of donkey sentences, whose pronouns resist a straightforward treatment as independently bound variables.

During the formal consolidation of these approaches, You Watanabe’s 1988 assignment-transition analysis established an equivalence between file updates and relational state changes for the existential-conjunctive fragment. Under corresponding accessibility conditions, the two representations produce the same admissible input-output assignment pairs. The result provided a direct translation between file-based analyses and the relational notation subsequently used in dynamic predicate systems.

Dynamic predicate logic, developed by Jeroen Groenendijk and Martin Stokhof, internalized context change within the interpretation of first-order formulas. Frank Veltman developed update semantics for modality and conditionals, treating their interpretation as operations on information states rather than as ordinary truth-functional composition.

Relational interpretation

In dynamic predicate logic, a formula denotes a binary relation between variable assignments. For assignments (g) and (h), the notation

[ \llbracket \varphi \rrbracket(g,h) ]

states that processing (\varphi) can transform (g) into (h). An atomic formula normally tests the input assignment without changing it:

[ \llbracket P(x) \rrbracket(g,h) \quad\text{iff}\quad g=h \text{ and } g(x)\in P. ]

Existential quantification changes the value assigned to a variable while preserving the values of the remaining variables:

[ \llbracket \exists x,\varphi \rrbracket(g,h) ]

holds when an assignment differing from (g) at most in the value of (x) can be connected to (h) by the relation denoted by (\varphi). The variable introduced by the existential can remain available outside its local syntactic constituent because its value persists in the output assignment.

Sequential composition supplies the dynamic interpretation of conjunction:

[ \llbracket \varphi \land \psi \rrbracket(g,h) ]

holds exactly when there is an intermediate assignment (k) such that

[ \llbracket \varphi \rrbracket(g,k) \quad\text{and}\quad \llbracket \psi \rrbracket(k,h). ]

This treatment explains why the discourse “A sailor owns a telescope, and she maintains it” permits both pronouns to depend on values introduced in the first clause. The first conjunct changes the assignment state by introducing suitable individuals, while the second conjunct places additional conditions on those same values.

Accessibility and subordinate domains

Not every introduced referent remains accessible. Dynamic negation ordinarily functions as a test on the current state rather than as an operation that exports newly assigned values. A schematic rule is

[ c[\neg\varphi]

{i\in c \mid {i}[\varphi]=\varnothing}. ]

The update retains an input possibility only when updating that possibility with (\varphi) fails. Referents introduced during the hypothetical evaluation of (\varphi) are therefore unavailable after the negation has been completed. This accounts for the contrast between a referent introduced in an asserted clause and one introduced in a negated clause.

Conditionals create a comparable asymmetry. In “If a researcher owns a telescope, she insures it,” the referents introduced in the antecedent are accessible within the consequent because the consequent is evaluated relative to antecedent-supporting states. Those referents do not ordinarily become accessible after the entire conditional, since the conditional does not assert that any researcher or telescope exists in the actual discourse state.

These patterns correspond to structural accessibility relations in discourse representation theory. A subordinate representation can inherit accessible referents from a containing representation, while its locally introduced referents remain confined unless the construction licenses their projection. The dynamic and representational formulations differ in machinery but encode closely related constraints.

Presupposition and information states

Dynamic semantics also supplies a framework for presupposition. A presuppositional expression can be modeled as imposing a definedness condition on its input context. If the required information is already present, the update proceeds; if it is absent, the update is undefined unless the context is expanded through accommodation.

For example, “The laboratory reopened” presupposes the existence of a contextually identifiable laboratory and conventionally indicates that it had previously ceased operating. A dynamic account distinguishes the information required before the sentence can be integrated from the information asserted by the completed update. This distinction permits presuppositions to interact compositionally with negation, conditionals, and quantified expressions.

The same general architecture applies to epistemic modality. An information state supports “must (\varphi)” when every live possibility satisfies (\varphi), whereas it supports “might (\varphi)” when at least one live possibility is compatible with (\varphi). Modal expressions can thus test the structure of an information state without necessarily reducing it in the manner of an ordinary assertion.

Relation to truth-conditional semantics

Dynamic semantics does not eliminate truth-conditional semantics. A sentence can retain ordinary truth conditions while receiving a richer interpretation as a context transformer. For a closed sentence (\varphi), truth in a model can often be recovered by examining whether the update succeeds from an appropriate initial state.

The distinction concerns the semantic type assigned to expressions. A static proposition has a type corresponding to a set of possible worlds, while a dynamic proposition has a type corresponding to an operation on information states. Under suitable assumptions, a static proposition can be lifted into the dynamic system by treating it as a test that removes incompatible possibilities. Not every dynamic operation reduces to such a test, because operations that introduce discourse referents alter the internal structure of the state.

This difference also separates dynamic semantics from a purely pragmatic account of cross-sentential interpretation. Pragmatic reasoning remains relevant to referent selection and contextual accommodation, but the persistence and inaccessibility of discourse referents are encoded in the compositional semantics itself. The framework therefore assigns formal significance to the order in which semantic material changes a context.

See also

  • Discourse representation theory, which models discourse interpretation through structured referents and accessibility relations.
  • Dynamic predicate logic, which interprets first-order formulas as relations between variable assignments.
  • Update semantics, which analyzes sentences as operations on informational states.
  • File change semantics, which represents indefinites and pronouns through evolving assignment files.
  • Formal semantics, which studies meaning through mathematically specified models and compositional interpretation.
  • Anaphora, which concerns dependencies between linguistic expressions and contextually available antecedents.
  • Donkey sentence, which exhibits pronoun dependencies not captured by elementary variable binding.
  • Common ground, which represents information treated as shared within a discourse context.