Lexical Semantics

Polysemy

One word, many related meanings — how a single lexical item branches across senses

Polysemy is the property of a single word having multiple related meanings derived from a common conceptual core. The English verb "run" — run a race, run a company, run a fever, run for office, the run of the river — exemplifies a single phonological form spanning fifty-plus distinct senses in the OED. Polysemy contrasts with homonymy ("bank" of a river vs. financial "bank"), where two unrelated meanings share a form by historical accident. The distinction was sharpened by Michel Bréal in his 1897 Essai de sémantique. Cognitive linguists (George Lakoff, Claudia Brugman, 1980s) modeled polysemy as a radial category — a prototypical sense plus metaphorical and metonymic extensions. Almost every high-frequency word in every language is polysemous; monosemy is the rare exception, mostly confined to technical jargon.

  • CoinedMichel Bréal, 1897 (Essai de sémantique)
  • ContrastPolysemy (related senses) vs. homonymy (unrelated)
  • OED senses of "set"430+ — largest in English
  • Radial category modelBrugman (1981); Lakoff (1987)
  • MechanismMetaphor, metonymy, specialization, generalization
  • UniversalityAll natural languages; rare in formal/technical registers

Interactive visualization

Press play, or step through manually. The visualization is yours to drive — try it before reading on.

Open visualization fullscreen ↗

Watch the 60-second explainer

A condensed visual walkthrough — narrated, captioned, under a minute.

Why polysemy matters

  • Lexicography. Sense division shapes every dictionary entry — splitters multiply, lumpers consolidate.
  • NLP. Word sense disambiguation drives translation, search, and question answering.
  • Cognitive science. Radial categories reveal how concepts extend metaphorically.
  • Language acquisition. Children master core senses early; figurative senses by adolescence.
  • Translation. Polysemy rarely aligns across languages — "run" maps to dozens of verbs in French.
  • Diachrony. Polysemy traces semantic change — extensions become primary senses over centuries.
  • Pragmatics. Speakers exploit polysemy for puns, double meanings, deliberate ambiguity.

Common misconceptions

  • Polysemy equals homonymy. Polysemy is related senses; homonymy is unrelated meanings sharing form.
  • Each sense is a separate word. Senses cluster around a core, not as discrete lexicons.
  • Context fully resolves ambiguity. Some senses remain genuinely indeterminate in usage.
  • Dictionaries list all senses. Speakers create new metonymic uses constantly; dictionaries lag.
  • One sense is "literal," others "figurative." Many extensions are conventionalized — no felt metaphor remains.
  • Polysemy is sloppy or imprecise. It is a productive cognitive resource, not a flaw of natural language.

Frequently asked questions

How do linguists distinguish polysemy from homonymy?

Three tests are standard. First, etymology — "bank" (riverbank) is from Old Norse bakki, while "bank" (financial) is from Italian banca; unrelated origins indicate homonymy. Second, semantic relatedness — "mouth" of a river vs. "mouth" of an animal share an embodied schema (opening, conduit), so polysemous. Third, the zeugma test — "He ran a marathon and a small business" is acceptable (polysemy); "She deposited a check at the bank she fished from" produces a clash (homonymy). Dictionaries list polysemes as numbered subentries under one headword and homonyms as separate entries (bank¹, bank²).

What is a radial category?

A model proposed by Claudia Brugman (1981, M.A. thesis on "over") and elaborated by George Lakoff in Women, Fire, and Dangerous Things (1987). The central or prototypical sense is surrounded by extensions linked through metaphor, image schemas, or metonymy. For "over": the central sense is spatial superposition ("the lamp over the table"); extensions include covering ("a tablecloth over the table"), excess ("over budget"), completion ("the show is over"), repetition ("do it over"), and authority ("the manager over the team"). Each extension is motivated, not arbitrary, but speakers do not derive senses on the fly — they are stored.

Why are frequent words more polysemous?

Zipf's law of meaning (G.K. Zipf, 1945) — the number of meanings of a word correlates with its frequency, roughly with the square root. High-frequency words ("get," "go," "make," "do," "have," "take") are recruited for novel uses because they are cheap to retrieve and broadly applicable. Specialized vocabulary ("photosynthesis," "phoneme") stays monosemous because it occupies a narrow conceptual niche. Children acquire polysemous senses gradually — Eve Clark (1993) documented English-speaking children mastering core senses by age four and metaphorical extensions through adolescence.

What are metonymy and metaphor as polysemy mechanisms?

Metaphor maps structure across domains — "grasp an idea" extends manual prehension to comprehension. Metaphor extensions of "see" include understanding ("I see your point"), encountering ("see a doctor"), and dating ("seeing someone"). Metonymy uses contiguity — a part for a whole, container for contents, producer for product. "The White House announced" (institution for residents); "I'm reading Austen" (author for works); "nice wheels" (part for vehicle). Both mechanisms are productive — speakers create new metonymic uses on the fly, some of which conventionalize into polysemy. Lakoff and Johnson's Metaphors We Live By (1980) catalogued hundreds of conceptual metaphors underlying English polysemy.

Is polysemy a problem for natural language processing?

Word sense disambiguation (WSD) was an open AI problem for decades. WordNet (George Miller, Princeton, 1985 onward) catalogues senses for English; Babel Net extends multilingually. Statistical methods (Yarowsky, 1995) and neural embeddings (BERT, 2018; contextualized embeddings) now resolve most disambiguation by leveraging context. But fine-grained sense distinctions still confuse models — humans disagree on 20-30% of WordNet sense annotations, suggesting inventories are too granular. Recent work uses substitutability or usage clusters rather than fixed sense lists.

Are polysemy patterns universal across languages?

Many polysemy patterns recur cross-linguistically. The body-part-to-spatial-relation extension ("head" of a list, "foot" of a mountain) appears in English, Mandarin, Yoruba, and Tzeltal. Eyes for understanding, hands for control, hearts for emotion are widespread but not universal — Japanese localizes emotion in the hara (belly), Aymara in the heart but with different metaphorical valences. Catherine Travis (2006) and Cliff Goddard's Natural Semantic Metalanguage program seek the universal core under language-specific patterns. The field of lexical typology (Maria Koptjevskaja-Tamm) maps systematic polysemy patterns across the world's languages.

What is regular polysemy?

A predictable, productive pattern where a class of words systematically displays the same sense relation. Animal-for-meat ("chicken," "lamb," "salmon" — but not "cow/beef" or "pig/pork," which retains Old French butcher terms after the Norman Conquest, 1066). Container-for-contents ("kettle is boiling," "the glass spilled"). Plant-for-fruit ("apple," "cherry"). James Pustejovsky's Generative Lexicon (1995) formalized regular polysemy with qualia structure — words carry typed roles (constitutive, formal, telic, agentive) generating senses by composition rather than enumeration. Irregular polysemy ("paper" — material, document, newspaper, academic article) requires listed senses.