Loading field map
How can the meaning of a sentence be modeled with the precision of logic? That question drives formal semantics, a subfield that applies mathematical tools—set theory, model theory, and recursive functions—to natural language. Two commitments unify the enterprise: compositionality (the meaning of a complex expression is determined by the meanings of its parts and the rules that combine them) and truth-conditional analysis (to know a sentence's meaning is to know the conditions under which it would be true). Over the past century, a succession of frameworks has refined, challenged, and extended these commitments, each responding to pressures that earlier approaches left unresolved.
The story begins with Gottlob Frege's Fregean Compositional Semantics (1892–present). Frege introduced the distinction between sense (Sinn) and reference (Bedeutung), arguing that a term's sense determines its reference and that the reference of a sentence is its truth value. Crucially, he insisted that the meaning of a sentence is built from the meanings of its parts in a rule-governed way—the principle of compositionality. This was a radical departure from earlier psychologistic or associative theories of meaning. Frege's framework, however, was limited to a logical language and did not provide a systematic account of truth itself.
That gap was filled by Tarskian Model-Theoretic Semantics (1935–present). Alfred Tarski defined truth for formal languages by specifying a recursive satisfaction relation between expressions and models (set-theoretic structures). A sentence is true if it is satisfied by all assignments of objects to variables. Tarski's approach gave formal semantics a rigorous method for linking language to the world, but it was designed for artificial languages and could not directly handle natural language's intensional contexts (e.g., belief reports, modal statements).
Rudolf Carnap extended Tarski's extensional framework into the intensional realm with Carnapian Intensional Semantics (1947–present). Carnap proposed that the meaning of an expression could be identified with its intension—a function from possible states of affairs to extensions. For example, the intension of a sentence is a function from possible worlds to truth values. This move preserved compositionality while accommodating necessity and other intensional phenomena. Carnap's work directly prepared the ground for the next major step.
Possible Worlds Semantics (1963–present), developed by Saul Kripke, Jaakko Hintikka, and others, turned Carnap's intensions into a fully worked-out model theory for modal logic. A possible world is a complete way things could be; necessity is truth in all worlds, possibility truth in some. This framework provided a clean semantics for modal operators and, later, for propositional attitudes. It coexisted with a different project: Davidsonian Truth-Conditional Semantics (1967–present). Donald Davidson proposed that a theory of meaning for a natural language could take the form of a Tarski-style truth theory that recursively assigns truth conditions to every sentence. Davidson insisted on staying extensional—no possible worlds, no intensions—and used the truth predicate as the central explanatory notion. His program narrowed the focus to the logical form of natural language sentences, especially action sentences and indirect discourse.
At the same time, Event Semantics (1967–present) emerged from Davidson's work on action sentences. Davidson argued that verbs introduce an implicit event variable, so "Brutus stabbed Caesar" is analyzed as "There was a stabbing of Caesar by Brutus." This addition of events to the semantic ontology allowed a compositional treatment of adverbial modification (e.g., "Brutus stabbed Caesar violently with a knife") that had resisted earlier approaches. Event semantics thus extended Davidsonian truth-conditional semantics by enriching the underlying ontology.
A more radical alternative appeared with Game-Theoretic Semantics (1968–present), proposed by Jaakko Hintikka. Instead of defining truth via a static model, game-theoretic semantics treats meaning as a two-player game between a verifier and a falsifier. A sentence is true if the verifier has a winning strategy. This procedural, interactive approach challenged the static model-theoretic picture, but it remained a niche framework, largely because its treatment of quantification and negation diverged from mainstream compositional methods.
Montague Semantics (1970–present) achieved a synthesis that reshaped the field. Richard Montague showed that natural language syntax and semantics could be paired in a precise, recursive fashion using the tools of intensional logic. He absorbed Carnap's intensions and Kripke's possible worlds into a unified system: every syntactic rule has a corresponding semantic rule, and the meaning of a sentence is its intension (a function from possible worlds to truth values). Montague's framework went beyond Davidson's by handling intensional contexts directly and by providing a systematic treatment of noun phrases, quantifiers, and anaphora. For a time, Montague Semantics became the dominant paradigm, setting the agenda for formal semantics in both philosophy and linguistics.
Even as Montague's approach flourished, researchers identified limitations and proposed extensions. Generalized Quantifier Theory (1981–present), developed by Jon Barwise, Robin Cooper, and others, extended Montague's treatment of quantifier phrases. Montague had analyzed "every man" as a property of properties; generalized quantifier theory formalized this insight using set-theoretic relations between sets (e.g., "every" denotes the subset relation). This framework connected directly to Tarskian model theory and provided a unified account of determiners like "most" and "few" that resist first-order analysis.
Alternative Semantics (1973–present), introduced by Mats Rooth, departed from the single-truth-value model of possible-worlds intensions. Rooth argued that focus-sensitive expressions (e.g., "only", "also") require a semantic value that is a set of alternatives—the ordinary meaning plus a set of contextually relevant alternatives. For example, "John only introduced BILL to Sue" contrasts Bill with other possible introducees. Alternative semantics thus added a new dimension to meaning without abandoning compositionality, coexisting with possible-worlds semantics as a specialized tool for focus and questions.
Situation Semantics (1981–present), proposed by Jon Barwise and John Perry, took a different tack. It incorporated many insights from possible-worlds semantics but adopted a more fine-grained ontology of situations—partial, spatiotemporally located parts of worlds. This allowed a treatment of perception reports and context-dependence that possible-worlds semantics struggled with. Situation semantics also influenced the development of dynamic approaches by emphasizing the role of information flow across discourse.
Dynamic Semantics (1981–present) challenged the static, sentence-level assumptions of Montague's framework. Hans Kamp's Discourse Representation Theory (DRT) and Irene Heim's File Change Semantics independently proposed that the meaning of a sentence is not its truth conditions but its potential to update a discourse context. An indefinite noun phrase like "a donkey" introduces a new discourse referent, which can then be picked up by a pronoun in a subsequent sentence. This dynamic view absorbed Montague's compositional machinery but transformed it: meaning became a relation between input and output contexts, not a function from worlds to truth values. Dynamic semantics proved especially fruitful for anaphora, presupposition, and cross-sentential phenomena.
Inquisitive Semantics (2009–present), developed by Jeroen Groenendijk, Floris Roelofsen, and others, extended the dynamic tradition to questions and inquisitive content. In inquisitive semantics, the meaning of a sentence is a set of possibilities (the information states that resolve the issue it raises). Declaratives propose to eliminate possibilities; interrogatives propose to partition the space of possibilities. This framework unified declarative and interrogative meaning within a single compositional system, building on both dynamic semantics and earlier work on questions in alternative semantics.
Today, no single framework dominates formal semantics. Instead, researchers draw on a toolkit of approaches, each suited to different phenomena. Possible-worlds semantics remains the default for modality and propositional attitudes. Dynamic semantics is standard for anaphora and discourse phenomena. Alternative semantics is the go-to framework for focus and questions, while inquisitive semantics offers a unified treatment of declaratives and interrogatives. Generalized quantifier theory is foundational for quantifier interpretation. Event semantics is widely adopted for adverbial modification and aspect. Montague's compositional method is still the backbone of most formal analyses, even when the specific ontology or dynamic machinery differs.
What the leading frameworks agree on is the centrality of compositionality and the use of model-theoretic tools. They disagree on the nature of semantic values: static intensions vs. context-change potentials vs. sets of alternatives. They also disagree on the ontology—whether to admit possible worlds, situations, events, or discourse referents as primitive. This pluralism is not a sign of fragmentation but of a mature field that tailors its formal apparatus to the explanatory demands of different linguistic domains. The history of formal semantics is thus a story of successive refinements, each framework preserving what worked in its predecessors while addressing new puzzles about how language conveys meaning.