What is a ‘word’?: different views

For speakers of some languages, the ‘word’ is a robustly intuitive notion. But it seems that no matter what the language, we have a hard time providing an exact definition that encompasses all and only those entities that our intuition tells us are words (see, e.g., Anderson 1985b: 153–4). This means that the concept ‘word’ is nothing if not elusive, and suggests that perhaps there is no concept of word that is universally applicable. Indeed, if there is no cross-linguistic, or universal psycholinguistic evidence for the existence of the word, then we may well doubt the validity of the word as a primitive natural language construct. It could a priori be the case that there is really no such thing in absolute terms as the ‘word’, and that it is just an artifactual linguistic construct that happens to coincide with salient units intermediate between morphemes and phrases that happen to appear in many of the world's languages.

There is another reason why the possibility that the ‘word’ is a derived rather than primitive construct may occur to us: words are definable using several disparate linguistic criteria. For some of these criteria considered in isolation, the label ‘word’ seems strangely inappropriate, since words so defined seem overly abstract, with nothing very ‘word-like’ about them. Let us take a look at these criteria to see if any of them are closer than others in providing an accurate portrayal of ‘word’.