Loading field map
Formal semantics is the branch of the philosophy of language that studies meaning by treating natural language as a system that can be described with the precise tools of logic and mathematics. Its central project is to explain how the meanings of complex expressions—sentences, phrases, and even discourses—are determined by the meanings of their parts and the rules by which those parts are combined. The guiding assumption is that natural language meaning is compositional: the meaning of a whole is a function of the meanings of its constituents and their syntactic mode of combination. This assumption, often called the principle of compositionality, is not merely a working hypothesis but the foundational commitment that distinguishes formal semantics from other approaches to meaning.
The field’s core questions are deceptively simple. What is the meaning of a declarative sentence? What does it mean for a sentence to be true or false, and how does that depend on the world? How do expressions like "every," "some," "the," and "most" contribute to truth conditions? How do pronouns, tenses, and modal auxiliaries like "might" and "must" get their values? How do we account for the fact that a sentence can be meaningful even when it is false, or that two sentences can say the same thing in different ways? Beneath these questions lies a deeper philosophical concern: what kind of thing is meaning itself? Is it a relation between language and the world, a set of conditions under which a sentence is true, a rule for updating information, or something else entirely? Formal semantics answers these questions by constructing explicit models—mathematical structures that assign denotations to expressions and compute the denotations of larger expressions from smaller ones.
The stakes are philosophical as well as technical. If natural language can be given a rigorous semantic theory, then many traditional philosophical problems become tractable. Questions about truth, reference, necessity, and the relationship between language and thought can be reformulated with precision. Conversely, the limits of formal methods reveal something about the nature of language itself. If certain phenomena resist formalization, that is a fact about human cognition and communication that any adequate theory of meaning must respect. Formal semantics thus sits at the intersection of philosophy, linguistics, and logic, and its practitioners often move freely among these disciplines.
Formal semantics grew out of the tradition of logical analysis that began in the late nineteenth century with the work of Gottlob Frege. Frege’s revolutionary insight was that the meaning of a sentence could be understood as its truth value—whether it is true or false—and that this truth value is determined by the references of its parts. He introduced the distinction between sense and reference: the reference of an expression is the object it picks out, while its sense is the mode of presentation of that object. The sense of a sentence, on Frege’s view, is a thought, and the reference is its truth value. This distinction allowed him to explain why "the morning star" and "the evening star" can both refer to Venus while differing in meaning: they present the same object in different ways.
Frege also developed the first formal language adequate for expressing the logical structure of sentences. His concept-script, or Begriffsschrift, introduced quantifiers and variables, making it possible to represent sentences like "Every philosopher is mortal" as a relation between concepts rather than as a simple subject-predicate structure. This was a decisive break with traditional grammar, which treated such sentences as having the same logical form as "Socrates is mortal." Frege showed that the logical form of a quantified sentence is radically different: it involves a second-order claim about the extension of a concept, not a claim about a particular object.
The next major step came from Bertrand Russell, who applied Frege’s tools to a range of philosophical problems. Russell’s theory of descriptions, developed in 1905, showed how phrases like "the present king of France" could be analyzed without assuming that they refer to an object. According to Russell, a definite description is not a referring expression at all but a complex quantificational phrase. The sentence "The present king of France is bald" is analyzed as a conjunction of three claims: there is a present king of France, there is only one such king, and that king is bald. Since the first claim is false, the whole sentence is false, not meaningless. This analysis demonstrated that formal logic could dissolve apparent philosophical puzzles about reference and existence.
Russell’s work, along with that of Frege and the early Ludwig Wittgenstein, established the idea that the logical form of a sentence might differ from its grammatical form. This idea became the basis of logical atomism, the view that the world consists of atomic facts and that language, at its deepest level, mirrors this structure. While logical atomism as a metaphysical doctrine did not survive, its methodological legacy did: the practice of analyzing natural language sentences by translating them into a formal language with a precise semantics.
The modern discipline of formal semantics took shape in the 1960s and 1970s, when philosophers and linguists began applying the tools of model theory to natural language. Model theory, developed by Alfred Tarski and others in the 1930s, provides a way of defining truth for a formal language. A model is a set of objects together with assignments of objects to names, sets of objects to predicates, and relations to relational expressions. Truth is then defined recursively: a sentence is true in a model if the relevant objects stand in the relevant relations. Tarski’s work showed that the concept of truth could be rigorously defined for formal languages, and it provided the template for defining truth for natural language.
The key figure in this transition was Richard Montague, who in the late 1960s and early 1970s argued that there is no important theoretical difference between natural languages and the formal languages of logic. Montague’s grammar, developed in a series of papers collected as Formal Philosophy, treated English as a formal language whose syntax and semantics could be described in tandem. His approach had three components: a syntactic analysis that assigned phrase structures to sentences, a semantic analysis that assigned denotations to expressions, and a translation procedure that linked the two. The semantics was intensional, meaning that it could handle not just truth and reference but also modality, tense, and propositional attitudes.
Montague’s most influential contribution was his treatment of noun phrases as generalized quantifiers. Instead of treating "every philosopher" as a name-like expression that refers to a set, he treated it as a function that takes a predicate and returns a truth value. "Every philosopher is mortal" is true if the set of philosophers is a subset of the set of mortals. This analysis unified the treatment of proper names, definite descriptions, and quantified phrases, and it provided a natural way of handling the semantics of "most," "few," and other quantifiers that resist first-order analysis. Montague’s work also introduced the lambda calculus as a tool for semantic composition, allowing complex meanings to be built up from simpler ones in a systematic way.
Montague’s program was not without its critics. Some philosophers objected that his semantics was too complex and too remote from actual linguistic practice. Others noted that his treatment of intensional contexts, such as "John believes that the morning star is the evening star," required a heavy apparatus of possible worlds and individual concepts. Nevertheless, Montague’s work established the basic architecture of formal semantics: a recursive syntax, a model-theoretic semantics, and a compositional mapping between them. This architecture remains the default framework for the field, even as subsequent developments have modified and extended it.
One of the central problems for any semantic theory is the treatment of expressions that do not simply refer to objects in the actual world. Modal expressions like "necessarily" and "possibly," temporal expressions like "will" and "was," and attitude verbs like "believes" and "knows" all seem to require a semantics that goes beyond truth and reference. The standard solution, developed by Saul Kripke and others in the 1950s and 1960s, is the framework of possible worlds. A possible world is a complete way the world could be; the actual world is one among many. A sentence like "It is necessary that $2+2=4$" is true if the embedded sentence is true in all possible worlds. A sentence like "It is possible that it rains tomorrow" is true if the embedded sentence is true in at least one possible world.
Possible worlds semantics provided a powerful tool for formal semantics, but it also raised philosophical questions. What are possible worlds? Are they concrete entities, as David Lewis argued, or are they abstract representations, as Kripke preferred? The debate over the nature of possible worlds is not merely metaphysical; it affects the explanatory power of the semantics. If possible worlds are primitive, then the semantics explains necessity in terms of quantification over worlds, but it does not explain what makes a world possible. If possible worlds are constructed from language or from properties, then the semantics may be circular.
A related issue is the treatment of propositional attitudes. The sentence "John believes that the morning star is the evening star" seems to attribute to John a belief about a relation between two objects. But if the morning star and the evening star are the same object, then the belief seems to be about a trivial identity. The problem is that belief contexts are not truth-functional: substituting co-referring terms can change the truth value of the whole sentence. Possible worlds semantics handles this by treating the object of belief as a proposition, which is a set of possible worlds. "John believes that p" is true if the proposition expressed by p is among the worlds compatible with John’s beliefs. But this analysis has its own problems. If propositions are sets of worlds, then any two sentences true in the same worlds express the same proposition. This makes it impossible to distinguish between necessarily equivalent sentences, such as "$2+2=4$" and "All bachelors are unmarried," even though a person might believe one and not the other. This problem, known as the problem of logical omniscience, remains a live issue in the field.
By the 1980s, a new approach emerged that challenged the static, truth-conditional picture of meaning. The problem was that many natural language expressions seem to have their meaning only in relation to a context of utterance. Indexicals like "I," "here," and "now" refer to the speaker, the place, and the time of the utterance. Demonstratives like "this" and "that" depend on the speaker’s gestures or intentions. And the interpretation of pronouns often depends on the preceding discourse. A purely truth-conditional semantics, which assigns meanings to sentences in isolation, seems unable to handle these phenomena.
The response was the development of dynamic semantics, which treats the meaning of a sentence not as a proposition but as a change in the context. In dynamic semantics, a sentence is interpreted as a function from contexts to contexts. The context is typically modeled as a set of possibilities—the information available to the participants in the conversation. A sentence like "A man walked in" updates the context by adding the information that there is a man who walked in. A subsequent sentence like "He was wearing a hat" then picks out that man as its referent. This approach, developed by Irene Heim and Hans Kamp in the early 1980s, provided a unified account of anaphora, presupposition, and the way information accumulates in discourse.
Dynamic semantics was not a rejection of compositionality but a reinterpretation of it. The meaning of a sentence is still a function of the meanings of its parts, but the parts are now understood as context-change potentials rather than as static denotations. This shift had important consequences. It made it possible to handle donkey sentences like "Every farmer who owns a donkey beats it," which had resisted truth-conditional analysis. In dynamic semantics, the indefinite "a donkey" introduces a discourse referent that can be picked up by the pronoun "it," even though the pronoun is outside the scope of the quantifier. The analysis was controversial, and alternative static treatments were proposed, but the dynamic framework demonstrated that the boundaries of formal semantics could be extended to phenomena that had previously seemed beyond its reach.
Contemporary formal semantics is a diverse field, unified by its commitment to rigorous, explicit models of meaning but divided over the proper form of those models. The Montagovian tradition remains influential, particularly in its treatment of quantification and intensionality. But it has been supplemented by a range of other approaches. Type-theoretic semantics, which builds on the simply typed lambda calculus, provides a flexible framework for handling a wide range of syntactic constructions. Event semantics, developed by Donald Davidson and Terence Parsons, treats verbs as denoting events, allowing for the analysis of adverbial modification and the progressive aspect. Plural semantics extends the framework to deal with plural noun phrases and collective predicates. And the study of presupposition and implicature has become a major subfield in its own right, drawing on the work of Paul Grice and Robert Stalnaker.
One of the most significant developments of recent decades is the integration of formal semantics with the study of pragmatics. The traditional division of labor assigned semantics to the study of truth conditions and pragmatics to the study of speaker meaning. This division has become increasingly untenable. Many expressions, such as "some" and "or," seem to have a core semantic meaning that is enriched in context. The sentence "I ate some of the cookies" semantically entails that I ate at least one cookie, but in most contexts it implicates that I did not eat all of them. The question of how to draw the line between what is said and what is implicated remains a central topic of debate. Some theorists, following Grice, treat implicatures as pragmatic inferences that are computed on the basis of the literal meaning. Others, following the relevance theory of Dan Sperber and Deirdre Wilson, argue that the literal meaning is itself underdetermined and that context plays a role in determining even the truth-conditional content of an utterance.
Another important development is the growing attention to the relationship between formal semantics and cognitive science. The question of whether the structures posited by formal semantics correspond to anything in the minds of speakers is a matter of ongoing debate. Some researchers, particularly those influenced by Noam Chomsky’s generative grammar, treat formal semantics as a description of a mental faculty. Others, influenced by the tradition of model-theoretic semantics, treat it as a description of the relationship between language and the world, independent of psychology. This disagreement is not merely philosophical; it affects the kinds of evidence that are considered relevant. Cognitive approaches tend to rely on experimental data about how speakers process sentences, while model-theoretic approaches tend to rely on intuitions about truth and falsity.
The field also faces a number of unresolved problems. The treatment of vagueness, as in "tall" or "bald," remains controversial. The semantics of conditionals, particularly counterfactuals, is still a matter of active research. The relationship between semantics and syntax is not fully understood, and different frameworks make different assumptions about how the two are connected. And the question of whether natural language is fully compositional, or whether there are irreducible idioms and constructions, continues to be debated.
Despite these open questions, formal semantics has achieved a remarkable degree of consensus on its basic methodology. The idea that meaning can be studied by constructing explicit models and testing them against intuitions about truth and inference is now widely accepted, not just in philosophy but in linguistics as well. The field has produced a rich body of results that have transformed our understanding of quantification, anaphora, tense, modality, and a host of other phenomena. And it has provided a common language in which philosophers, linguists, and logicians can discuss questions about meaning with a precision that would have been unimaginable a century ago. The result is not a finished theory but a living research program, one that continues to expand the boundaries of what can be said about what we say.