Loading field map
Semantics is the branch of linguistics concerned with meaning. More precisely, it is the systematic study of how meaning is encoded in natural language and how it is recovered by speakers and hearers. While the word "semantics" is sometimes used loosely in everyday speech to refer to the interpretation of any sign or symbol, within linguistics it names a specific technical discipline: the analysis of the meanings of words, phrases, sentences, and larger stretches of discourse, and the principles by which those meanings compose.
The central puzzle of semantics is that meaning is both stable and creative. A finite vocabulary and a finite set of grammatical rules allow speakers to produce and understand an indefinitely large number of novel sentences. Any adequate theory of meaning must explain how this is possible: how the meaning of a complex expression is determined by the meanings of its parts and the way they are combined. This is known as the compositionality problem, and it has shaped the field more than any other single concern.
Before any analysis can begin, semantics must decide what kind of thing "meaning" is. Three broad answers have historically been available, and they continue to structure the field.
The first answer is referential: the meaning of an expression is what it points to in the world. The meaning of "cat" is the class of cats; the meaning of "the cat on the mat" is a particular situation. This view is intuitive but runs into immediate difficulties. Two expressions can refer to the same thing yet differ in meaning—"the morning star" and "the evening star" both refer to the planet Venus, but they do not mean the same thing, since a speaker can believe one statement without believing the other. Referential semantics alone cannot capture this distinction.
The second answer is mentalist: meaning is a concept or idea in the mind. The meaning of "cat" is the mental representation of a cat. This view respects the intuition that meaning has something to do with thought, but it faces the problem of privacy. If meanings are private mental entities, it is hard to explain how communication works at all, since no two people can directly compare their concepts.
The third answer is use-based: the meaning of an expression is its use in the language. This view, associated with the later philosophy of Ludwig Wittgenstein, holds that meaning is not a hidden entity but a pattern of behavior. To know the meaning of "cat" is to know how to use it correctly in sentences and situations. This approach is attractive for its concreteness, but it struggles to explain how we understand novel sentences we have never heard before, since their use has not yet been established.
Modern linguistic semantics does not choose one of these answers wholesale. Instead, it typically adopts a hybrid: meaning is modeled as a relation between language and the world (reference), but the relation is mediated by mental representations and constrained by patterns of use. The dominant formal tradition treats meaning as a function from contexts to truth conditions, a position that owes much to the referential tradition but incorporates mental and use-based elements.
The modern discipline of semantics emerged in the mid-twentieth century, but its intellectual roots lie in earlier traditions. In the nineteenth and early twentieth centuries, meaning was studied primarily within philology and historical linguistics, where the focus was on how word meanings change over time—narrowing, broadening, and shifting. This tradition, sometimes called historical semantics, asked questions like "How did the English word 'meat' come to mean animal flesh rather than food in general?" It was an empirical, descriptive enterprise, but it had no general theory of what meaning is or how it works.
A more theoretical impulse came from philosophy, particularly from the tradition of analytic philosophy and the work of Gottlob Frege at the turn of the twentieth century. Frege distinguished between sense and reference: the reference of an expression is the thing it picks out, while the sense is the mode of presentation of that thing. "The morning star" and "the evening star" have the same reference but different senses. Frege also insisted that the meaning of a sentence is its truth value—whether it is true or false—and that the meanings of parts combine to determine the truth value of the whole. This insight, later developed by Alfred Tarski into a formal theory of truth, became the backbone of formal semantics.
The decisive step came in the 1960s and 1970s, when linguists, most prominently Richard Montague, applied the tools of formal logic to natural language. Montague's central claim was that there is no important theoretical difference between natural languages and the artificial languages of logic. Both can be described by the same kind of formal grammar, and both can be given a precise semantics in terms of truth conditions. This program, known as Montague grammar or model-theoretic semantics, treated a sentence's meaning as the conditions under which it would be true in a model—a formal representation of a possible world. The meaning of a word was a function that could combine with the meanings of other words to yield the truth conditions of sentences.
This approach had a profound effect on the field. It made semantics rigorous and testable: a semantic theory could now be judged by whether it correctly predicted the truth conditions of an infinite set of sentences. It also connected linguistics to logic and the philosophy of language in a productive way. The central tool became compositional semantics: the systematic rules by which the meanings of parts combine. For example, the meaning of "every cat sleeps" is built from the meaning of "every," "cat," and "sleeps" by a rule that says "every" takes a property (being a cat) and another property (sleeping) and asserts that the first property is included in the second.
Within the field, a useful distinction is drawn between two levels of analysis. Lexical semantics studies the meanings of individual words and the systematic relations among them. It asks how word meanings are structured, how they relate to one another, and how they change. Compositional semantics studies how word meanings combine to form the meanings of phrases and sentences. The two are interdependent: composition rules operate on lexical meanings, and lexical meanings are only fully understood in the contexts of the sentences they can appear in.
Lexical semantics has developed its own set of analytic tools. One important tradition analyzes word meanings in terms of semantic features—primitive components of meaning that combine to form more complex concepts. For example, "man" might be analyzed as [+human, +male, +adult], while "woman" is [+human, −male, +adult]. This approach, associated with structuralist linguistics, is useful for capturing systematic relations like synonymy, antonymy, and hyponymy (the "is a kind of" relation). But it faces the problem of defining the primitives: what are the ultimate features, and how many are there? No agreed-upon list has ever been produced.
A more recent approach to lexical semantics, associated with cognitive linguistics, treats word meanings as prototypes rather than sets of necessary and sufficient conditions. On this view, the meaning of "bird" is not a definition but a typical example—a robin is a more central bird than a penguin. This approach captures the graded nature of many categories and explains why some borderline cases are hard to classify. However, it has difficulty accounting for the precise, all-or-nothing judgments that compositional semantics requires.
Compositional semantics, by contrast, has been dominated by the formal tradition. Its central achievement is the development of a type-theoretic framework in which every expression is assigned a semantic type, and combination rules specify how types can combine. The basic types are those of entities (things) and truth values (true or false). A proper name like "Alice" is an entity; a sentence like "Alice sleeps" is a truth value; a verb like "sleeps" is a function from entities to truth values. This framework handles a remarkable range of phenomena, including quantification ("every," "some," "no"), anaphora (pronouns and their antecedents), and intensional contexts (sentences about beliefs and desires, where the truth of the embedded clause is not simply a matter of the facts).
A major challenge to formal semantics came from the recognition that much of what speakers communicate is not literally said. The sentence "Can you pass the salt?" is, in its literal meaning, a question about ability. But in ordinary conversation, it is almost always understood as a request. This gap between what is said and what is meant is the subject matter of pragmatics, the study of language use. The philosopher H. P. Grice proposed that communication is governed by a cooperative principle and a set of maxims—be informative, be truthful, be relevant, be clear—and that speakers can deliberately flout these maxims to convey meanings beyond the literal content. The sentence "It's cold in here" might be used to mean "Please close the window," not because the sentence means that, but because the hearer infers the request from the context and the assumption that the speaker is being cooperative.
The relationship between semantics and pragmatics is one of the most contested issues in the field. On one view, semantics deals with the stable, context-independent meaning of sentences, while pragmatics deals with the context-dependent inferences that hearers draw. On another view, the boundary is much fuzzier. Many expressions—indexicals like "I" and "here," demonstratives like "this" and "that," and gradable adjectives like "tall"—cannot be assigned a determinate meaning without context. "I am tall" is true for some speakers and false for others, depending on who says it and what the comparison class is. Some theorists argue that such context-dependence is pervasive and that the traditional division between semantics and pragmatics cannot be maintained.
This debate has given rise to a family of approaches sometimes called contextualist, which hold that the truth conditions of many sentences are not fully determined by the words alone but require pragmatic enrichment. The rival position, sometimes called minimalist, holds that semantics should aim for a minimal, context-independent core of meaning, leaving all contextual adjustment to pragmatics. The dispute is not merely terminological: it affects how one analyzes specific constructions, such as "John is ready" (ready for what?) or "I've had breakfast" (today? ever?). These are live research questions, not settled doctrines.
Not all of semantics is formal. A significant tradition, often grouped under the label cognitive linguistics, rejects the assumption that meaning can be captured by truth conditions and formal composition. Instead, it treats meaning as a matter of conceptualization: the meaning of an expression is the mental experience it evokes. On this view, language is not a system for describing an objective world but a tool for organizing and communicating subjective experience.
Cognitive semantics emphasizes the role of image schemas—recurring patterns of bodily experience, such as containment, path, and force—in structuring meaning. Prepositions like "in," "out," "through," and "over" are analyzed not as arbitrary formal relations but as expressions of these embodied schemas. The approach also emphasizes metaphor as a central mechanism of meaning: abstract concepts like time, love, and argument are understood through systematic metaphorical mappings from concrete domains. The claim is not that metaphors are decorative but that they are constitutive of how we think and speak.
A related but distinct tradition is usage-based semantics, associated with functional linguistics and construction grammar. This approach holds that the meaning of an expression is not a fixed mental representation but a generalization over the situations in which the expression is used. Meaning is learned from exposure to actual instances of language, and it is stored in the form of constructions—pairings of form and meaning that range from single words to complex syntactic patterns. On this view, there is no sharp boundary between lexicon and grammar: both are collections of constructions of varying complexity.
These approaches differ from formal semantics not only in their methods but in their explanatory goals. Formal semantics aims to explain how speakers can produce and understand novel sentences; cognitive and usage-based approaches aim to explain how meanings are learned, stored, and processed. The two traditions often talk past each other, but they are not necessarily incompatible. A complete theory of meaning would need to account both for the systematic composition of novel expressions and for the psychological and social processes by which meanings are acquired and used.
Contemporary semantics is a pluralistic field. The formal tradition remains the dominant paradigm in many linguistics departments, particularly in North America and Europe, and it has produced a sophisticated body of work on quantification, tense and aspect, modality, presupposition, and anaphora. Its methods have been extended beyond the sentence to discourse semantics, which studies how the meaning of a text or conversation is built up across multiple sentences, and to dynamic semantics, which treats the meaning of a sentence as its effect on the information state of the hearer rather than as a static truth condition.
At the same time, the cognitive and usage-based traditions have grown in influence, particularly in the study of lexical meaning, grammaticalization, and language acquisition. The rise of corpus linguistics—the analysis of large collections of naturally occurring language—has provided empirical evidence for usage-based claims and has made it possible to study meaning in ways that were not feasible with introspective methods alone.
A notable recent development is the increasing interaction between semantics and computational linguistics. The task of building systems that can understand and generate natural language has forced researchers to make semantic theories precise enough to be implemented. This has led to new interest in formal semantics among computationalists, and to new challenges for the field: how to handle vagueness, context-dependence, and the massive ambiguity of natural language in a way that is computationally tractable.
The field is also marked by unresolved foundational questions. The nature of the semantics-pragmatics boundary, the role of context in determining truth conditions, the relationship between meaning and reference, and the psychological reality of semantic representations are all actively debated. What is not in dispute is the central insight that has guided the field since its modern founding: meaning is systematic, and it can be studied with the same rigor as other aspects of language. The task of semantics is to make that systematicity explicit.