A unit of expression which has universal intuitive recognition by native-speakers, in both spoken and written language. However, there are several difficulties in arriving at a consistent use of the term in relation to other categories of linguistic description, and in the comparison of languages of different structural types. These problems relate mainly to word identification and definition. They include, for example, decisions over word boundaries, as well as decisions over status.

Three main senses of ‘word’ are usually distinguished (though terminology varies):

(a) Words are the physically definable units which one encounters in a stretch of writing (bounded by spaces) or speech (where identification is more difficult, but where there may be phonological clues to identify boundaries, such as a pause, or juncture features). ‘Word’ in this sense is often referred to as the orthographic word (for writing) or the phonological word (for speech). A neutral term often used to subsume both is word form.

(b) There is a more abstract sense, referring to the common factor underlying the set of forms which are plainly variants of the same unit, such as walk, walks, walking, walked. The ‘underlying’ word unit is often referred to as a lexeme. Lexemes are the units of vocabulary, and as such would be listed in a dictionary.

(c) This then leaves the need for a comparably abstract unit to be set up to show how words work in the grammar of a language, and ‘word’, without qualification, is usually reserved for this role (alternatively, one may spell out this implication, referring to ‘morphemic / morphosyntactic / grammatical’ words, though the latter has an alternative sense). A word, then, is a grammatical unit, of the same theoretical kind as morpheme and sentence. In a hierarchical model of analysis, sentences (clauses, etc.) consist of words, and words consist of morphemes (minimally, one free morpheme). Word-order refers to the sequential arrangement of words in a language. Languages are sometimes classified in terms of whether their word-order is relatively ‘free’ (as in Latin) or ‘fixed’ (as in English).

 Several general subclassifications of words have been proposed, such as the distinction between variable and invariable types, grammatical (or function) words v. lexical words, closed-class v. open-class words, empty v. full words. At a more specific level, word-classes can be established, by analysing the various grammatical, semantic and phonological properties displayed by the words in a language, and grouping words into classes on the basis of formal similarities (e.g. their inflections and distribution).