Much will be concerned with how adjectives can be derived from nouns, nouns from verbs, and so on. It is important therefore that terms for word classes such as ‘adjective’, ‘noun’ and ‘verb’ should be properly understood. (What I have just called word classes are the same as what in traditional terminology are called parts of speech and what many contemporary linguists call lexical categories.) Readers who are confident that they can recognize a noun or a verb when they see one may feel entitled to skip to the next topic. On the other hand, I suspect that many such confident readers think that the word class to which a lexeme belongs is mainly determined by its meaning. That belief is incorrect. If you feel tempted by it.

In school, you may once have been told that verbs are ‘doing words’, while nouns are ‘thing words’ and adjectives ‘describing words’. The trouble with these meaning-based definitions is that, if one takes them seriously, they require us to lump together lexemes whose grammatical behavior is quite different, and distinguish between ones whose grammatical behavior is similar. Consider again the lexeme PERFORM, which looks like a prototypical ‘doing word’, denoting something that actors and musicians do. The lexeme PERFORMANCE denotes the same activity, surely. Does that mean that PERFORM and PERFORMANCE belong to the same word class? That can hardly be right, since they occur in such different syntactic contexts, and since their inflectional behavior is so different: PERFORMANCE has the two forms performance (singular) and performances (plural), while PERFORM has the four forms performs, performed, performing and perform. In fact, as we have seen, PERFORMANCE is a noun and PERFORM is a verb. This classification can be made, solely on the basis of their syntactic and inflectional behavior, with no appeal to meaning – and indeed meaning may be positively misleading, since a performance is not obviously a ‘thing’.

Compare now the lexemes PERFORM and RESEMBLE. Is the latter a ‘doing word’ too? That seems scarcely appropriate. Resembling, one may think, hardly counts as an activity. To say that (for example) my great uncle William resembles a giraffe is not to report some action of his, but rather to describe him. Should we then lump RESEMBLE in with other supposed ‘describing words’ – adjectives such as TALL and INTERESTING? Again, this meaning-motivated conclusion falls foul of syntactic and inflectional evidence. These adjectives have comparative and superlative forms (taller, tallest) or phrasal substitutes for them (more interesting, most interesting); on the other hand, RESEMBLE has a set of forms (resembles, resembled, resembling and resemble) exactly parallel to the forms of PERFORM, and used in broadly parallel syntactic contexts. So to identify verbs as ‘doing words’ risks misleading us into neglect of the syntactic and inflectional parallels that justify classifying not only PERFORM but also RESEMBLE as a verb.

Does that mean, then, that a lexeme cannot have both noun forms (singular and plural) and verb forms (past, third person singular present, and so on)? If part of identifying a lexeme is identifying what word class it belongs to, then that must be true – but trivially so, because it amounts to decreeing that a root that can carry verbal suffixes such as -ed and -ing as well as the noun plural suffix -s must belong to two lexemes, not one. The more interesting question, then, is: do such roots exist? The answer is certainly yes. For example, HOPE and FEAR have both noun forms (her hope/fear for the future) and verb forms (she hoped/feared that it would rain). Other similarly ambivalent words are WISH, DESIRE, FATHER (a verb in He fathered seven children), and COOK. Does this mean that the concept ‘word class’, as I have used it, is too vague or inconsistent to be useful?

The answer is no, for two reasons. The first involves the proportion of our noun–verb vocabulary that is ambivalent in this way. Although numerous, it is still heavily outnumbered by the proportion that is either purely noun-like in its grammatical behavior (e.g. DOOR, SISTER, DESK, JOY) or purely verb-like (e.g. HEAR, SPEAK, WRITE, BELIEVE). Admittedly, one can imagine a language in which a far higher proportion of the vocabulary is ambivalent in the way we are discussing, and in respect of such a language one might well argue that many or most lexemes did not belong to identifiable word classes. Such claims have in fact been made in relation to some languages in the Austronesian family, which contains (for example) Malay, Tagalog, and the languages of Polynesia, as well as some native languages of western Canada and the US Pacific coast. Even there, however, it seems generally necessary to distinguish nominal (i.e. ‘nouny’) and verbal syntactic structure, despite the fact that the class of lexemes that can occur in each type of structure is almost the same.

A second kind of reason has to do with English in particular. Let us compare HOPE and FEAR as verbs with other verbs that can be followed by that-clauses, as in (1):

(1) a. She stated that it would rain.

b. She knew that it would rain.

c. She denied that it would rain.

d. She admitted that it would rain.

e. She acknowledged that it would rain.

For all of these sentences we can identify a nominal counterpart, that is a counterpart of the form her … that it would rain:

(2) a. her statement that it would rain

b. her knowledge that it would rain

c. her denial that it would rain

d. her admission that it would rain

e. her acknowledgement that it would rain

What is striking about the nouns in (2) is that they all involve a suffix added to the basic form of the verb in (1) (possibly with some other phonological change, as in knowledge and admission). There are few verb–noun pairs that one can use in the contexts of (1) and (2) such that the basic and suffixed forms are the other way round, the noun supplying the base and the verb being derived from it by means of a suffix. In morphological terms, therefore, it makes sense to say that the verbal construction in (1) is basic, the nominal construction in (2) being derived from it. But this has implications for HOPE and FEAR as well. If we look only at (3) and (4), we have no basis for deciding whether these lexemes are basically nominal or basically verbal:

(3) a. She hoped that it would rain.

b. She feared that it would rain.

(4) a. her hope that it would rain

b. her fear that it would rain

However, as soon as we notice that (3) and (4) are parallel to (1) and (2) respectively, we have a ground for concluding that HOPE and FEAR are basically verbal. The nominal contexts of (4) are parallel to those of (2), where the nouns are clearly derived from verbs; so it makes sense to say that the nouns HOPE and FEAR in (4) are derived from verbs too, even though they carry no affix.

The notion that derivation can occur without any overt change in shape may at first seem strange. Some linguists have accordingly decided that HOPE and FEAR as nouns, are really ‘zero-derived’, carrying a phonologically empty and therefore unpronounceable ‘zero suffix’: HOPE-∅, FEAR-∅. Others have preferred to say that one of the processes available in derivational morphology is conversion, whereby a lexeme belonging to one class can simply be ‘converted’ to another, without any overt change in shape. We do not need to decide here which is the better style of analysis, though I will generally refer to the phenomenon as ‘conversion’. Either way, these ambivalent words present the problem of determining which word class the basic form belongs to. Sometimes, as with HOPE and FEAR a decisive argument involving parallels with affixed lexemes can be found. Sometimes, despite the risks already mentioned of relying on meaning as a criterion, the basic meaning seems clearly appropriate to one word class rather than another; for example, few would deny that, even though FATHER can function as a verb, it is the noun (as in my father) that is more basic.