We saw that it is possible for a linguistic item to be a basic building-block of syntax – that is, an item that is clearly not itself a sentence or a phrase – and yet to have a meaning that is predictable. We saw, in other words, that characteristic 2. does not necessarily entail characteristic 1. We will see that characteristic 1. does not necessarily entail characteristic 2.: that is, something that is clearly larger than a word (being composed of two or more words) may nevertheless have a meaning that is not entirely predictable from the meanings of the words that compose it.

Consider these two sentences from the point of view of a learner of English who is familiar with the usual meanings of the words expenditure, note and tab:

(6) I keep notes on all my expenditure.

(7) I keep tabs on all my expenditure.

Will this learner be able to interpret both these sentences accurately? The answer, surely, is no. Sentence (6) presents no problem; the learner should be able to interpret it correctly as meaning ‘I write down a record of everything I spend’. But faced with sentence (7), on the basis of the usual meaning of tab, the learner is likely to be puzzled. Does it mean something like ‘I attach small flaps to all the notes and coins that I spend’? Or perhaps ‘I tear off small pieces from the paper money that I spend, and keep them’? Neither interpretation makes much sense! Native speakers of English, however, will have no difficulty with (7). They will instinctively interpret keep tabs on as a single unit, meaning ‘pay close attention to’ or ‘monitor carefully’. Thus, keep tabs on, although it consists of three words, functions as a single unit semantically, its meaning not being predictable from that of these three words individually. In technical terms, keep tabs on is an idiom. Even though it is not a word, it will appear in any dictionary that takes seriously the task of listing semantic idiosyncrasies, probably under the headword tab.

Idioms are enormously various in length, structure and function. Keep tabs on behaves rather like a verb, as do take a shine to ‘become attracted to’, raise Cain ‘create a disturbance’, have a chip on one’s shoulder ‘be resentful’, and kick the bucket ‘die’. Many idioms behave more like nouns, as the following pair of sentences illustrates:

(8) The interrogation took a long time because the suspect kept introducing irrelevant arguments.

(9) The interrogation took a long time because the suspect kept introducing red herrings.

Again, a learner of English might be puzzled by (9): did the suspect keep pulling fish from his pocket? A native speaker, however, will know that red herring is an idiom meaning ‘irrelevant argument’, so that (8) and (9) mean the same thing. Other noun-like idioms are white elephant ‘unwanted object’, dark horse ‘competitor whose strength is unknown’, Aunt Sally ‘target of mockery’.

In most of the idioms that we have looked at so far, all the individual words (tabs, shine, bucket, elephant etc.) have a literal or non-idiomatic meaning in other contexts. Even in raise Cain, the fact that Cain is spelled with a capital letter hints at a reference to the elder son of Adam, who, according to biblical legend, murdered his younger brother Abel. However, there are also words that never occur except in an idiomatic context. Consider these examples:

(10) My aunt took pains to get the answer right.

(11) My aunt took part in the conversation.

(12) My aunt took offence at the suggestion.

(13) My aunt took umbrage at the suggestion.

(10), (11) and (12), take pains, take part and take offence all deserve to be called idioms, because they are multi-word items whose meaning is not fully predictable from their component words. (To a learner of English, (11) might seem to imply that my aunt was present during only part of the conversation, and (12) might suggest that she committed an offence.) If so, then presumably we should say the same of (13), containing the phrase take umbrage at. The difference between (13) and the others, however, is that umbrage does not appear anywhere except in this phrase (in my usage, at least). This restriction means that it would not really be sufficient for a dictionary to list umbrage as a noun meaning something like ‘annoyance’; rather, what needs to be listed is the whole phrase. Similarly, the word cahoots exists only in the phrase in cahoots with ‘in collusion with’, and it is the whole phrase which deserves to be lexically listed, as an idiom.

Akin to idioms, but distinguishable from them, are phrases in which individual words have collocationally restricted meanings. Consider the following phrases:

(14) white milk

(15) white coffee

(16) white noise

(17) white man

Semantically, these phrases are by no means totally idiosyncratic: they denote a kind of milk, coffee, noise and man, respectively. Nevertheless, in a broad sense they may count as idiomatic, because the meaning that white has in them is not its usual meaning; rather, when collocated with milk, coffee, noise and man respectively, it has the meanings ‘yellow’, ‘brown (with milk)’ (at least in British usage), ‘containing many frequencies with about equal amplitude’, and ‘belonging to an ethnic group whose members’ skin color is typically pinkish or pale brown’.

If a typical idiom is a phrase, then a word with a collocationally restricted meaning is smaller than a typical idiom. That provokes the question whether there are linguistic items with unpredictable meanings that are larger than phrases – specifically, that constitute whole sentences. The answer is yes: many proverbs fall into this category. A proverb is a traditional saying, syntactically a sentence, whose conventional interpretation differs from what is suggested by the literal meaning of the words it contains. Examples are:

(18) Too many cooks spoil the broth.

‘Having too may people involved in a task makes it harder to complete.’

(19) A stitch in time saves nine.

‘Anticipating a future problem and taking care to avoid it is less troublesome in the long run than responding to the problem after it has arisen.’

(20) It’s no use crying over spilt milk.

‘After an accident one should look to the future, rather than waste time wishing the accident had not happened.’

Here again, it is useful to distinguish between predictability and motivation. The relationship between the literal meaning and the conventional interpretation of these proverbs is not totally arbitrary. Rather, the conventional interpretation is motivated in the sense that it arises through metaphorical extension of the literal meaning. For example, spilling milk is one kind of accident, but in the proverb at (20) it is used metaphorically to stand for any accident. However, idioms are still unpredictable in the sense of being conventional; for example, one cannot freely invent a new idiom such as ‘It’s no use crying over a broken plate’, even though its metaphorical meaning may be just as clear as that of (20).

If idioms are listed in dictionaries (usually via one of the words that they contain), should proverbs be listed too? As it happens, ordinary dictionaries do not usually list proverbs, because they are conventionally regarded as belonging not to the vocabulary of a language but to its usage (a rather vague term for kinds of linguistic convention that lie outside grammar). For present purposes, what is important about proverbs is that they constitute a further example of a linguistic unit whose use and meaning are in some degree unpredictable, but which is larger than a word.