1 Definition and scope
1.1 Core concept
A multiword expression is a sequence of two or more words that is treated by speakers as a single unit in meaning, use, or form. Such expressions may be fully fixed or allow limited variation, but they are recognized as conventional combinations rather than newly assembled phrases each time they occur. The category includes idioms, collocations, phrasal verbs, formulaic routines, and many other recurrent combinations.
1.2 Distinction from single-word lexical items
Unlike a single-word lexical item, a multiword expression is made up of separate orthographic or phonological words. Even so, it may behave like one lexical unit in interpretation and storage. For example, an expression can have a specialized sense, a restricted pattern of use, or a strong association between its parts that makes it function differently from a freely composed phrase.
1.3 Boundaries and related terms
The boundaries of the category are not always sharp. Linguists differ on whether some recurring sequences should count as multiword expressions, constructions, or simply frequent phrases. The field therefore uses overlapping terms to describe expressions that are fixed, semi-fixed, or only partially predictable.
1.3.1 Phrase
A phrase is any group of words forming a grammatical unit. Many phrases are freely generated and fully compositional, so not all phrases qualify as multiword expressions. The latter term is narrower, emphasizing conventionality, stored status, or reduced predictability.
1.3.2 Fixed expression
A fixed expression is a conventional sequence whose wording is highly stable. Its parts are usually not replaced, reordered, or expanded without affecting acceptability. Fixed expressions often include idioms, sayings, and routine formulas.
1.3.3 Lexical bundle
A lexical bundle is a recurrent sequence identified mainly by its frequency in corpus data. It may be structurally complete or incomplete and may not have idiomatic meaning. Lexical bundles are especially important in discourse analysis and language teaching because they show patterns of habitual use.
1.4 Criteria for identification
Researchers use several criteria to decide whether a sequence is a multiword expression. No single criterion is sufficient in all cases, so identification usually relies on a combination of semantic, syntactic, lexical, and statistical evidence.
1.4.1 Semantic unity
Semantic unity means that the expression conveys a meaning that is experienced as a single concept or conventional sense. The overall interpretation may be transparent, partly transparent, or highly idiomatic, but the sequence functions as one meaningful package.
1.4.2 Syntactic fixedness
Syntactic fixedness refers to limited flexibility in structure. A multiword expression may resist changes in word order, insertion of extra words, or alteration of grammatical relations. This resistance is often a sign that the sequence is stored and retrieved as a unit.
1.4.3 Lexical fixedness
Lexical fixedness involves restrictions on the substitution of one word for another. In many expressions, one or more elements are strongly preferred or obligatory. This property is common in collocations and idioms, where only certain lexical choices sound natural.
1.4.4 Frequency and conventionality
Frequent recurrence and social convention are central indicators of multiword status. An expression that appears often in a speech community becomes easier to recognize, learn, and process. Conventional use also helps stabilize form and meaning over time.
2 Types of multiword expressions
2.1 Idioms
Idioms are expressions whose meaning cannot be fully predicted from the literal meanings of their parts. They often have figurative senses, cultural associations, or historical origins that are not transparent to new speakers. Idioms can range from opaque to partly understandable.
2.1.1 Fully idiomatic expressions
Fully idiomatic expressions have meanings that are not recoverable from the individual words. Their interpretation must usually be learned as a whole, since literal analysis offers little guidance. Such expressions are often highly conventional and resistant to modification.
2.1.2 Semi-idiomatic expressions
Semi-idiomatic expressions combine literal and figurative meaning. One part may retain its ordinary sense while the overall phrase carries a specialized or conventional interpretation. These expressions are common in everyday language because they balance familiarity with expressive force.
2.2 Collocations
Collocations are words that tend to occur together more often than would be expected by chance. Their meaning is usually transparent, but the combination feels natural to native speakers. Typical collocations are important in both fluent speech and careful writing.
2.2.1 Strong collocations
Strong collocations have a very restricted set of preferred partners. A word may regularly co-occur with only one or a few others, making the combination especially salient. These pairings are often learned as ready-made units.
2.2.2 Weak collocations
Weak collocations allow more lexical freedom, yet still show a noticeable preference for certain combinations. The association is less rigid than in strong collocations, but conventional patterns still guide usage. They are often easier for learners to overlook because their meaning seems ordinary.
2.3 Phrasal verbs
Phrasal verbs consist of a verb combined with a particle or preposition to form a unit with a specialized meaning. They are especially common in English and may be literal, metaphorical, or highly idiomatic. Their behavior often includes variable word order and register sensitivity.
2.3.1 Particle verbs
Particle verbs combine a verb with an adverbial particle such as up, out, or off. The particle can alter aspect, direction, or meaning in ways that are not always predictable. Some particle verbs are separable, while others remain fixed in order.
2.3.2 Prepositional verbs
Prepositional verbs consist of a verb followed by a preposition that is required by the construction. The preposition is not merely an optional adjunct but part of the lexical pattern. These verbs are often learned as whole combinations because the choice of preposition is constrained.
2.4 Compounds and compound-like phrases
Compounds and compound-like phrases are combinations that function like single lexical items in naming or classification. They may be written solidly, with a hyphen, or as separate words. Their main feature is that the combination has a stable lexical identity.
2.4.1 Open compounds
Open compounds are written as separate words but behave as a unit in meaning or reference. Many everyday terms fall into this category. Their status shows that orthography alone does not determine lexical unity.
2.4.2 Hyphenated forms
Hyphenated forms link two or more elements with a hyphen to indicate close association. They often serve as modifiers or specialized labels. In practice, hyphenation may signal an intermediate stage between free phrase and established compound.
2.5 Formulaic expressions
Formulaic expressions are conventional utterances used in recurring social or discursive situations. They include set greetings, leave-takings, acknowledgments, and routine comments. Because they are tied to common interactions, they are often learned early and used automatically.
2.5.1 Greetings and routines
Greetings and routines are fixed or semi-fixed expressions used in predictable communicative settings. Examples include introductions, thanks, apologies, and farewells. Their value lies less in new information than in managing social interaction.
2.5.2 Discourse markers
Discourse markers help organize speech, signal transitions, or express stance. They may be short phrases with little propositional content, yet they play an important role in coherence and interaction. Their frequent use makes them especially characteristic of spoken language.
2.6 Binomials and trinomials
Binomials and trinomials are coordinated expressions containing two or three elements in a preferred order. They often sound fixed because the sequence is conventional and memorable. Many have rhythmic, semantic, or rhetorical patterns that encourage repetition.
2.6.1 Coordinated pairs
Coordinated pairs are two-part expressions joined by conjunctions or linked by conventional sequencing. Some pairs show a stable order, such as a customary first and second element. Their predictability contributes to their status as stored phrases.
2.6.2 Fixed triplets
Fixed triplets contain three coordinated or rhythmically linked elements. They are often used in rhetoric, storytelling, and legal or ceremonial language. The three-part structure can enhance memorability and emphasis.
3 Structural properties
3.1 Syntactic patterns
Multiword expressions occur in many grammatical shapes. Some resemble noun phrases, others behave like verbs, and many function as prepositional or adverbial units. Their syntax often reflects a balance between ordinary grammar and lexicalized patterning.
3.1.1 Noun phrases
Noun-phrase expressions are among the most common types, especially in naming and classification. They may include compounds, titles, technical terms, or recurring descriptive phrases. Their internal structure can be relatively fixed even when the phrase functions as a single noun-like unit.
3.1.2 Verb phrases
Verb-phrase expressions often include phrasal verbs, verbal idioms, and formulaic sequences with a verbal core. They may govern specific complements or particles. Many show restrictions that distinguish them from freely formed verb combinations.
3.1.3 Prepositional phrases
Prepositional-phrase expressions frequently serve as discourse formulas, idiomatic locutions, or fixed adverbial units. Their prepositions may be semantically extended or lexically obligatory. Such phrases often contribute nuance rather than direct referential content.
3.2 Word order constraints
Many multiword expressions permit little variation in the order of their components. Reversing elements can produce awkwardness or a different meaning altogether. Stable ordering is one of the most visible signs that an expression is conventionalized.
3.3 Morphological variation
Some expressions allow inflection or agreement while preserving their identity. A verb may change tense, a noun may be pluralized, or an adjective may vary in degree. Nevertheless, the expression remains recognizable because its core pattern is stable.
3.4 Internal modification
Internal modification refers to the insertion or alteration of material inside an expression. Some multiword expressions tolerate little interference, while others permit limited expansion. The degree of flexibility is often tied to how fixed or idiomatic the expression is.
3.4.1 Adjective insertion
Adjective insertion may be possible in expressions with a clear syntactic frame, especially when the phrase is less idiomatic. In highly fixed expressions, however, added modifiers can disrupt naturalness. This difference helps distinguish flexible constructions from rigid ones.
3.4.2 Inflectional changes
Inflectional changes include shifts in tense, number, aspect, or comparison. Many expressions allow such changes without losing identity, but some resist them because the exact wording is part of the conventional form. Variation often depends on the degree of lexicalization.
3.5 Grammaticality and acceptability
An expression can be grammatically possible yet still sound unnatural if it departs from conventional usage. Acceptability therefore depends not only on abstract grammar but also on phraseological expectation. Native-like fluency often requires knowledge of these restrictions.
4 Semantic properties
4.1 Literalness and idiomaticity
Multiword expressions range from fully literal combinations to highly idiomatic ones. Literal expressions can be understood by direct composition, while idiomatic ones rely on conventionalized meanings. Many expressions lie between these extremes and combine both patterns.
4.2 Compositionality
Compositionality is the extent to which an expression’s meaning can be derived from the meanings of its parts and their arrangement. Some expressions are highly compositional, whereas others are only partly so. This property is central to the study of phrase meaning and lexical storage.
4.2.1 Fully compositional expressions
Fully compositional expressions have meanings that follow straightforwardly from their components. Even if they are frequent, they do not require special semantic interpretation. Their multiword status may rest more on lexical choice than on semantic opacity.
4.2.2 Partially compositional expressions
Partially compositional expressions contain at least one element whose contribution is specialized, extended, or metaphorical. The whole remains partly interpretable, but conventional usage shapes its final meaning. Such expressions are common in everyday language and technical discourse.
4.3 Polysemy and semantic shift
Many multiword expressions develop multiple related senses over time. A phrase may begin literally and later acquire figurative or specialized uses. Semantic shift can make an expression more compact, more expressive, or more domain-specific.
4.4 Pragmatic meaning
Pragmatic meaning concerns what an expression does in context beyond its basic reference. Multiword expressions often signal politeness, familiarity, irony, emphasis, or social alignment. Their impact depends on setting, participants, and communicative goals.
4.4.1 Speaker stance
Speaker stance refers to the attitudes or evaluations conveyed through an expression. A formula, interjection, or idiomatic phrase may express certainty, caution, approval, or skepticism. These meanings are often indirect but socially important.
4.4.2 Social convention
Social convention shapes which expressions are appropriate in particular situations. Some phrases are tied to ritual, etiquette, or repeated interaction. Their force comes from shared expectations rather than novel semantic content.
5 Lexical and phraseological behavior
5.1 Lexical fixedness
Lexical fixedness is the tendency for a phrase to preserve specific word choices. In many multiword expressions, replacing one component with a near synonym makes the phrase sound incorrect or unnatural. This property is a major reason such expressions are memorized as wholes.
5.2 Constructions and patterning
Some expressions are best understood as constructions, meaning recurring form-meaning pairings with slots for variable elements. The pattern itself carries information in addition to the words used. This perspective highlights regularity across many similar expressions.
5.3 Register and stylistic value
Multiword expressions often carry strong register associations. Some sound formal, archaic, colloquial, humorous, or technical. Writers and speakers use these associations to shape tone and audience response.
5.4 Regional and dialectal variation
Different speech communities may prefer different expressions for the same function or meaning. Regional variation can affect vocabulary, pronunciation, and conventional phrasing. A sequence familiar in one variety of a language may be rare or unknown in another.
5.5 Lexical creativity and variation
Speakers sometimes adapt established expressions for humor, emphasis, or stylistic effect. Such creativity may preserve the original pattern while replacing one element, or it may deliberately twist a familiar phrase. Recognition depends on shared knowledge of the original form.
6 Acquisition and use
6.1 First-language acquisition
Children often acquire many multiword expressions as unanalyzed chunks before fully understanding their internal structure. Frequent routines and caregiver formulas are especially early acquisitions. Over time, learners may begin to notice the parts and generalize from them.
6.2 Second-language learning
For second-language learners, multiword expressions are often difficult because they require knowledge of conventional combination, not just vocabulary and grammar. Mastery involves learning which words typically go together and how much variation is acceptable. These patterns are essential for natural-sounding language.
6.2.1 Memorization and chunking
Memorization helps learners store recurring sequences as chunks. Chunking reduces cognitive load by treating a phrase as a single retrieval unit. This strategy can improve fluency and accuracy when used with appropriate input.
6.2.2 Common learner difficulties
Learners often struggle with collocations, idioms, article use, prepositions, and phrasal verbs. Errors may arise from direct translation, partial knowledge, or overgeneralization. Because many expressions are only partly predictable, explicit attention is often helpful.
6.3 Formulaic language in fluency
Formulaic language supports fluent speaking and writing by providing ready-made building blocks. It allows the speaker to plan larger stretches of discourse while relying on familiar sequences for continuity. This can improve speed, coherence, and perceived naturalness.
6.4 Processing in speech and writing
Multiword expressions are often processed more quickly than novel combinations because they are familiar and expected. In comprehension, recognition can be facilitated by frequency and context. In production, stored phrases may reduce planning effort and help maintain rhythm.
7 Description and representation
7.1 Dictionary treatment
Dictionaries handle multiword expressions in several ways, depending on frequency, transparency, and user needs. Some are treated as independent entries, while others are listed within broader headwords. Good dictionary treatment balances economy with discoverability.
7.1.1 Headwording
Headwording refers to the decision to give an expression its own entry. This is common for idioms, fixed phrases, and widely used collocations. The goal is to make conventional expressions easy to locate and interpret.
7.1.2 Cross-referencing
Cross-referencing connects related expressions, variants, or component words. It helps users move between a phrase and its parts or between similar phraseological items. This is especially useful when an expression can be searched under more than one element.
7.2 Corpus linguistics
Corpus linguistics provides large bodies of real language data for studying multiword expressions. Researchers use corpora to identify recurrent sequences, measure their frequency, and examine their contexts. This approach has greatly expanded phraseological research.
7.2.1 Frequency analysis
Frequency analysis identifies combinations that recur unusually often. High frequency alone does not prove idiomaticity, but it does indicate conventionality. Statistical measures help distinguish ordinary co-occurrence from meaningful association.
7.2.2 Concordance methods
Concordance methods display an expression in many contexts so that patterns of use can be observed. By comparing lines of data, researchers can see variation, grammatical behavior, and pragmatic function. Concordances are especially valuable for studying collocations and formulaic sequences.
7.3 Annotation and tagging
Annotation and tagging add linguistic labels to corpus data, such as part of speech, syntactic role, or phrase boundary. These labels make it easier to search for and classify multiword expressions. Careful annotation supports both manual analysis and automated processing.
7.4 Machine processing
Computational methods are used to detect, classify, and interpret multiword expressions in text and speech. Because such expressions often resist purely compositional analysis, they are challenging for language technology. Progress in this area has improved many applications.
7.4.1 Automatic extraction
Automatic extraction seeks to find candidate expressions from large datasets without prior manual listing. Systems rely on patterns of recurrence, association, and context. The output usually requires human review because some candidates are accidental co-occurrences.
7.4.2 Statistical identification
Statistical identification uses measures such as mutual association, dispersion, and frequency to detect probable expressions. These methods are useful for ranking candidates and finding hidden patterns. They work best when combined with linguistic criteria.
7.4.3 Natural language processing applications
Natural language processing uses knowledge of multiword expressions in parsing, machine translation, information retrieval, and text generation. Recognizing these units can improve meaning interpretation and reduce errors caused by word-by-word analysis. Their treatment remains an active area of research.
8 Theoretical perspectives
8.1 Phraseology
Phraseology studies fixed and semi-fixed combinations as a core part of language structure. It emphasizes conventionality, recurrence, and restricted variation. The field has contributed many descriptive categories for multiword expressions.
8.2 Construction grammar
Construction grammar treats recurring form-meaning pairings as constructions, whether they are single words or larger patterns. From this viewpoint, many multiword expressions are not exceptions but ordinary grammatical units. The approach highlights learned patterns and inherited structure.
8.3 Generative approaches
Generative approaches often distinguish between regular rule-based phrase formation and stored lexical items. Multiword expressions are important because they challenge a strictly compositional model. They show that some meanings and structures are best explained as lexicalized exceptions or fixed patterns.
8.4 Cognitive linguistics
Cognitive linguistics emphasizes conceptualization, metaphor, categorization, and memory in the understanding of multiword expressions. Many idioms and phrasal patterns are analyzed as motivated by underlying image schemas or conceptual mappings. This perspective helps explain why some nonliteral expressions remain intelligible.
8.5 Usage-based approaches
Usage-based approaches view multiword expressions as products of repeated exposure and entrenchment. Frequent patterns become cognitively salient and easier to retrieve. On this account, expression, grammar, and lexicon develop through actual language use rather than separate modules.
9 Applications
9.1 Lexicography
Lexicography uses multiword expressions to improve dictionary coverage and user access. Recording established phrases helps describe meaning more accurately and reflects authentic usage. Dictionary makers must decide which expressions merit separate treatment.
9.2 Language teaching
Language teaching benefits from explicit attention to recurring phrases, collocations, and idioms. Teaching these units can strengthen fluency, accuracy, and comprehension. Instruction often focuses on noticing, practice, and contextualized retrieval.
9.3 Translation
Translation must account for multiword expressions because direct substitution often fails. Translators need to identify whether an expression should be rendered literally, idiomatically, or with a different conventional equivalent. This requires sensitivity to both meaning and usage.
9.4 Terminology work
Terminology work deals with specialized multiword terms in technical and professional fields. Many domains rely on stable phraseological units for precision and consistency. Standardizing these expressions supports communication across documents and institutions.
9.5 Natural language understanding
Natural language understanding systems must recognize multiword expressions to interpret text correctly. Without this capability, systems may misread idioms, phrasal verbs, or fixed routines. Accurate handling of such expressions improves semantic analysis and user interaction.