Polysemy (from Greek *poly-* 'many' and *sēma* 'sign') is the linguistic phenomenon whereby a single word or phrase possesses multiple related meanings. In semantics, polysemy is distinguished from homonymy (unrelated meanings) by the historical or conceptual connection between senses. For example, the English word "head" can denote a body part, a leader, the top of an object, or the foam on beer—all linked metaphorically or metonymically. Polysemy is pervasive in natural language, reflecting language economy and cognitive flexibility, and is studied through lexical semantics, cognitive linguistics, and computational lexicography.
1 Definition and scope
1.1 Distinction from homonymy
Polysemy involves multiple senses that share a common origin or conceptual link, whereas homonymy refers to distinct words that happen to share the same form (e.g., *bank* as a financial institution vs. *bank* of a river). In historical terms, polysemous senses typically derive from a single etymon, while homonyms often arise from separate etymological roots converging due to sound change or borrowing. Lexicographers use etymological evidence and semantic relatedness tests to make this distinction.
1.2 Distinction from monosemy and vagueness
Monosemy describes a word with a single, fixed meaning (e.g., *oxygen*). Vagueness, by contrast, occurs when a word's meaning is underspecified but not genuinely distinct (e.g., *neighbor* can refer to someone of either gender without being polysemous). Polysemy involves discrete, though related, senses that can be listed in a dictionary.
1.3 Polysemy vs. lexical ambiguity
Lexical ambiguity is a broader category that includes both polysemy and homonymy. In context, ambiguous forms are resolved by the surrounding discourse; polysemy often shows predictable patterns of sense extension, whereas homonymy is more arbitrary. Psycholinguistic studies demonstrate that readers process polysemous words faster than homonyms, suggesting a cognitive advantage for related senses.
2 Types of polysemy
2.1 Regular polysemy
Regular polysemy refers to systematic patterns of meaning extension that recur across many words in a language. These patterns are often governed by cognitive mechanisms like metaphor and metonymy.
2.1.1 Metonymic polysemy (e.g., container for contents)
Metonymy involves a shift based on contiguity. A typical example is the use of a container to refer to its contents: *dish* can mean a plate or the food served on it. Other common patterns include author-for-work (*read Shakespeare*), place-for-institution (*the White House announced*), and instrument-for-artist (*he plays the violin*).
2.1.2 Metaphorical polysemy (e.g., spatial to temporal)
Metaphor extends meaning by mapping from a concrete domain to an abstract one. Spatial prepositions like *in* and *on* acquire temporal meanings (*in the morning, on time*). Similarly, *bright* shifts from visual perception to intelligence (*bright idea*). Such mappings often follow systematic correspondences, such as time as space.
2.1.3 Categorical polysemy (e.g., animal for meat)
Certain nouns denote both an animal and its flesh used as food: *chicken, lamb, turkey*. This pattern involves a metonymic relationship between the source (the living animal) and the product (the meat). It is regular in English and many other languages, though not universal.
2.2 Irregular polysemy
Irregular polysemy arises from idiosyncratic historical or usage-based developments that do not follow a predictable pattern.
2.2.1 Historical drift
Over time, words can accumulate unrelated or loosely connected senses due to cultural change. For example, *mouse* originally meant only the rodent, but later acquired the sense of a computer peripheral through metaphorical extension. The shift is unique to the technology era and not part of a regular class.
2.2.2 Idiomatic specialization
Phrasal verbs or fixed expressions can create new senses that are not predictable from the component words. For instance, *break* in *break a habit* diverges from its core meaning. Such senses are often tied to specific constructions and may not extend to all uses of the word.
3 Cognitive and semantic mechanisms
3.1 Prototype theory and radial categories
Cognitive linguistics models polysemy using prototype theory: a word has a central, prototypical sense from which other senses radiate via family resemblances. For example, *crane* prototypically denotes a bird, with extended senses (construction equipment) linked by similarity in shape. These radial categories account for gradient relatedness among senses.
3.2 Metaphor and metonymy as sense extension
Metaphor and metonymy are primary drivers of polysemy. Metaphor transfers properties from a source domain (e.g., physical motion) to a target domain (e.g., emotion: *I'm feeling down*). Metonymy uses a salient part to stand for the whole or vice versa. Both mechanisms are grounded in embodied experience and can create systematic, cross-linguistic patterns.
3.3 Conceptual blending
Conceptual blending theory explains how novel senses arise by combining elements from two or more mental spaces. For instance, *surgeon* may be blended with *butcher* to create a derogatory sense (“a clumsy surgeon”). While less common for stable polysemy, blending accounts for creative, context-dependent extensions.
3.4 Semantic frames and domains
Frame semantics posits that word meanings are understood within structured background knowledge (frames). A word like *weekend* only makes sense relative to the concept of a seven-day work week. Polysemy arises when a word is used in different frames simultaneously, such as *free* (both “unfettered” and “costless”) anchored in distinct domains of liberty and commerce.
4 Polysemy in language change
4.1 Semantic bleaching and broadening
Semantic bleaching occurs when a word loses specific semantic content and becomes more general. For example, *thing* originally meant “assembly” but broadened to “entity.” Polysemy emerges as the narrowed sense coexists with the newer, broader one. Broadening can also happen in grammaticalization, where a full lexical item becomes a functional marker (e.g., *go* → future marker).
4.2 Sense narrowing and specialization
Conversely, a word may narrow from a general meaning to a specific one, leaving the original sense as a separate polyseme. Old English *deer* originally meant “animal” but later specialized to a particular species; the older sense survives in *deer* as a poetic archaism and in compounds like *reindeer*.
4.3 Reanalysis and folk etymology
Reanalysis can split a single form into multiple senses when speakers misinterpret the original structure. For instance, *hamburger* originally meant “from Hamburg,” but folk etymology led to the creation of *cheeseburger* and *veggie burger*, treating -*burger* as a morpheme. Similarly, *apron* was reanalyzed from *a napron* to *an apron*, but the older form survives in the borrowed sense of *napkin*.
5 Examples across languages
5.1 English (e.g., "run", "paper")
*Run* exemplifies high polysemy with over 600 recorded senses. Core meanings include physical motion (running), operation (running a machine), management (running a business), and extension (a run of luck). *Paper* can refer to material, a newspaper, an academic article, or wallpaper. The senses are linked through metonymy (material for product) and metaphor.
5.2 Chinese (e.g., 打 dǎ)
The Mandarin verb 打 *dǎ* (literally “strike”) has extended meanings covering actions like making a phone call (打电话), playing a game (打球), typing (打字), and buying (打车 “hail a taxi”). The pattern shows metaphorical extension from hitting to performing various manual and communicative acts, a relatively regular polysemy in Sinitic languages.
5.3 Romance languages (e.g., French "coup")
French *coup* (originally “blow”) has developed dozens of senses in fixed expressions: *coup de foudre* (love at first sight, literally “thunderbolt”), *coup d'État* (political takeover), *coup de grâce* (final blow), *coup de théâtre* (dramatic twist). Many metaphorical extensions rely on the core idea of a sudden, decisive action.
5.4 Polysemy in sign languages
Signed languages exhibit polysemy through iconic and spatial mappings. For instance, the American Sign Language sign for “tree” can also mean “forest” through reduplication. Metaphorical extensions are common: the sign for “heart” can mean “love” or “emotion.” Regular patterns, such as agent-instrument polysemy (e.g., “doctor” signed near the wrist as both the person and the act of healing), parallel spoken languages.
6 Computational and applied aspects
6.1 Word sense disambiguation
Word sense disambiguation (WSD) is the task of automatically identifying which sense of a polysemous word is used in a given context. It is critical for natural language understanding.
6.1.1 Supervised methods (e.g., sense-tagged corpora)
Supervised WSD uses human-annotated corpora (e.g., SemCor) where each occurrence of a word is labeled with a sense from a predefined inventory (e.g., WordNet). Classifiers such as support vector machines or neural networks learn contextual features (nearby words, syntactic patterns) to predict the correct sense. High accuracy is achieved for frequent words but data sparsity remains a challenge.
6.1.2 Unsupervised methods (e.g., clustering)
Unsupervised WSD clusters occurrences of a word based on distributional similarity without labeled data. Methods like k-means or hierarchical clustering group contexts into sense-like categories. These approaches are useful for discovering new senses in corpora but often yield clusters that do not align with dictionary definitions.
6.2 Lexical resources and ontologies (e.g., WordNet)
WordNet structures nouns, verbs, adjectives, and adverbs into synsets (sets of synonymous senses) linked by semantic relations (hypernymy, hyponymy, meronymy). Polysemy is represented by multiple synsets for the same word form. Similar resources include FrameNet (focus on frames) and BabelNet (multilingual). These databases support WSD, information retrieval, and knowledge extraction.
6.3 Polysemy in machine translation and NLP
Machine translation systems must resolve polysemy to produce accurate target-language equivalents. Statistical and neural models use contextual embeddings (e.g., BERT) to implicitly disambiguate. However, languages differ in how they lexicalize senses (e.g., English *river bank* vs. French *rive*), causing translation errors. Large language models mitigate this through massive context, but systematic evaluation shows residual failure with rare senses.
7 Philosophical and theoretical debates
7.1 Compositionality and sense enumeration
The principle of compositionality holds that the meaning of a complex expression is built from its parts. Polysemy poses a challenge because it seems to require multiple lexical entries for the same word. Sense enumerative approaches posit a list of discrete senses, while underspecification theories argue for a single core meaning that is modulated by context. The debate centers on whether polysemy is genuinely stored or generated online.
7.2 Core meaning vs. polysemy networks
Some linguists propose that every polysemous word has one basic or “core” meaning from which all others derive. Others advocate a network model, where senses are connected through multiple paths without a single central sense. Experimental evidence (e.g., reaction times, priming) supports the network view for some words but the core-meaning approach for others.
7.3 The role of context in sense selection
Context is crucial for resolving polysemy, but its exact role is debated. Interactionist theories claim that context actively shapes which sense is constructed, while selectionist theories assert that context merely chooses among pre-stored senses. Neurolinguistic studies show that both mechanisms may operate depending on the strength and predictability of a sense. Pragmatic factors (e.g., relevance theory) also contribute to how polysemy is understood in discourse.