The way I see it, a word could be conceptualized as different sets. I'll list them below.

Word as a singleton set: { to jump }

Word as a set of all its inflections: { to jump, jumps, jumped, jumping, have/has jumped }

Word as a set of all homographs with shared etymology / senses: { to jump (syn. of to hop), to jump (syn. of to attack) }

And then maybe there is a sense of word which involves a set that includes all the senses, as well as their respective inflections. So, which definition of word is correct within linguistics? If more than one of these are used, where are they used and why does there exist multiple senses of word? What are the theoretical reasonings behind the different senses?

  • 4
    The linguistic community has yet to find an adequate definition of ‘word’ (or ‘language’, for that matter) despite centuries of trying, so I doubt we’ll be able to settle it here. Commented Dec 18, 2021 at 9:42
  • 2
    For a more precise terminology, linguists avoid the term word and use more specific terms like word form, lexeme, or lemma. Commented Dec 18, 2021 at 19:46
  • Clearly related question fro the same OP: linguistics.stackexchange.com/questions/40356/… Commented Dec 18, 2021 at 19:47
  • 1
    Or doesn't have writing.
    – jlawler
    Commented Dec 19, 2021 at 17:59
  • 1
    I think there is nothing to do about it, I just wanted to create the link between the two questions such that interested readers can navigate between them. Commented Dec 20, 2021 at 21:19

2 Answers 2


Word, as noted, is not a clearly-defined term. In fact, P.H. Mathews, in his classic text Morphology, distinguishes three quite distinct usages of word:

  1. the Lexeme (the root, essentially the general meaning)
  2. the Word-Form (a combination of lexeme, phonology, and inflection in a particular case)
  3. the Phonological Word (the sounds that are involved)

I made up a diagram to illustrate the categories for my linguistics classes, using just the words lie and lay, which have quite confusing uses covering many contexts.

  • So, if I understand it correctly, Matthews classifies a lexemes by their meaning, meaning two words with the exact same spelling and pronunciation, even with similar meanings, are still different lexemes? Lie (i.e. recline) is different lexeme from lie (i.e. tell an untruth). In more nuanced situations, where the homonyms are senses of each other, they're still different lexemes? The word-form is the collection of a lexeme's inflectional forms? You are also showing the nominalization and verbalization connections on your diagram, so are they also included as a part of a word-form?
    – A. Kvåle
    Commented Dec 20, 2021 at 20:07
  • As for phonological word, it seems to me that as long as two words a homophones, they are the same phonological word. I don't see why this is a very relevant category within the diagram. I don't mean it seems irrelevant in it of itself, but to me, adding it in isolation makes the model seem unexhaustive. If the homophony relation is important enough to warrant a set of utterances that could be recognized as one of the designated meanings behind "word", why isn't there an equivalent set for homographs or synonyms?
    – A. Kvåle
    Commented Dec 20, 2021 at 20:11
  • Mathews says nothing about spelling. Linguists generally ignore English spelling because it doesn't represent the sounds of the word.
    – jlawler
    Commented Dec 20, 2021 at 23:03

I will elaborate on the notion of a phonological word, which seems to be somewhat in doubt. Typically, people think of words as being syntactic or morphological units, for example "cat" or "cats" are words – they are "syntactic terminals" in older theories of syntax. Words in that sense are also phonological words.

Clitics are special, in that they are syntactically individual units like lexical nouns and verbs are, but they are special phonology in that they seems to merge with a preceding or following (host) word. Thus in French, the object pronouns le, te are "syntactic separate words", but phonologically they attach to another word. Bantu languages often present mixed diagnoses for the status of clitics, so that in some senses they are outside of the "word" for rules that e.g. assign tone to the "word-final syllable", but they are attached to the word for rules that shorten word-final vowels. The concept of "phonological word" adds a necessary degree of freedom of analysis, corresponding to the fact that for some purposes clitics are "separate words" but for other words they are "the same word".

The puzzle of Northern Norwegian V2 can be partially solved by appeal to the notion of "phonological word". The verb must generally be in "second position", and will follow an initial (extracted) WH-word, for example ka sa dokker? "what did you pl. say", korsan går det "how does it go?". However, a monosyllabic WH-word and a monosyllabic pronoun merge into one, and can precede the verb, e.g. ka du sa "what did you say?", as though ka+du is one word.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.