Consider two similar sentences, such as (1) and (2):

(1) Edward sang the solos at the concert.

(2) The solos at the concert were sung by Edward.

Comparing these sentences, we find it natural to try to identify the respects in which they resemble one another and the respects in which they differ. We are likely to say that their lexical content is the same (they exploit the same lexical items), but they differ in that (2) is the passive sentence corresponding to the active sentence at (1). However, we are not inclined to describe (1) and (2) as ‘the same sentence’, in any sense. The expression ‘two forms of the same sentence’ has no application for us, whether as ordinary language users or, speaking more technically, as linguists. Probably this is because uttering or understanding a sentence is not usually a matter of recalling a single stored item from the memory – an item with which the sentence can be compared and judged ‘the same’. However, for present purposes what matters is simply the fact that (1) and (2) are not ‘the same sentence’, not the reasons for this fact.

Consider by contrast the following two word forms:

(3) sang

(4) sung

We feel these to be related also, but their relationship is different from that between (1) and (2). There is a clear sense in which, even as non-linguists, we feel them to be ‘the same word’. A dictionary will not assign to them two separate entries – or, more precisely, its entries for both sang and sung will simply refer the reader to the entry for sing. In the technical terminology, sang and sung are both word forms by means of which, in appropriate contexts, the lexeme SING is expressed. So there is an area of grammar, namely inflectional morphology, where it makes sense to talk of different forms of the same item. Consequently the processes that distinguish the word forms of a lexeme (processes of affixation, vowel change or whatever) differ in a fundamental respect from those that distinguish between sentences such as (1) and (2): they relate not different grammatical items but different forms of one item.

As well as being forms of one lexeme, sang and sung are also forms of one lexical item, we expect any English verb to have a past tense form and a perfect participle form, so it is not appropriate to record their existence by means of separate dictionary entries for these two forms of every verb. This is so even when their shapes (the word forms that express these grammatical words) need to be recorded because they are irregular; for this irregularity can be noted, where necessary, under a verb’s single dictionary entry. But, in the processes that relate these word forms, there is nothing that precludes them from being used to relate forms of distinct lexical items too. The kind of vowel change that relates sang to sing, and the kind of suffixation that relates performed to perform, do not come labelled ‘not to be used in relating distinct lexical items’. And these morphological processes are indeed used in English for this purpose, as in song (a distinct lexical item from sing) and performance (a distinct item from perform).

The existence of phrasal and sentential idioms shows that lexical items can perfectly well be formed by means of syntactic processes, whereby grammatical words are combined. But such word combinations are likely to be longer than the products of morphological processes such as affixation. Moreover, just by virtue of not being words, idioms are likely to less versatile syntactically than words are – that is, to be less convenient to fit into a wide variety of sentence types. So two factors, brevity and versatility, are likely to favor the morphological method over the syntactic method for creating lexical items. That being so, the considerable overlap between lexemes and lexical items becomes more readily understandable, and hence also the tendency to blur the distinction between them by calling them both ‘words’.

The account just offered in terms of English presupposes that inflectional morphology has a kind of priority over derivational. The notion ‘different word forms belonging to the same word’ is peculiar to inflectional morphology, and it is thus in inflectional morphology that processes for relating such word forms play their central role, even though these processes are available for exploitation elsewhere. It is only fair, in an introductory work such as this, to warn that this view of the status of derivational morphology relative to inflectional is not shared by all linguists. But that is not surprising, given what I said about the lack of any consensus on reasons for the overlap between ‘words’ as grammatical items and as lexical items.