¹ For criteria of ‘grammaticalness’, see especially Noam Chomsky, Syntactic structures 15 ff. ('s-Gravenhage, 1957).

² See M. A. K. Halliday, Categories of the Theory of Grammar §3.1 and §5.1 (to appear in Word). The present paper owes much to the stimulus of Halliday's remarks on grammar and lexis, especially §§2.1, 6.3, 7.3-4, 8.1, and in private discussion. I take this opportunity of expressing my thanks also to two other colleagues, J. M. Sinclair and J. P. Thorne, for helpful comments on an earlier draft of this paper, which was presented in February 1961 at a staff seminar at the School of Applied Linguistics, Edinburgh University.

³ For the term ‘collocation’, see J. R. Firth, Papers in linguistics 1984-1951 194 ff. (Oxford, 1957), and Studies in linguistic analysis 11 (Oxford, 1957).

⁴ A formal indication of this being the case is that they would collocate no better or worse with one word than with another. And if, by the imposition of a grammatical restriction, we were told that plint was to be assessed as a noun, there would still, for example, be no adjectives which we felt to be more appropriate or less appropriate than any others to qualify it. Cf. footnote 10.

⁵ An interesting lexical study could be made of the shifts of meaning undergone by words in fairy tales. An important point would be the very limited group of words so affected, all the rest remaining more or less stable and ‘normal’.

⁶ I may of course encounter a hitherto unfamiliar form, e.g. in a technical work, and decide that it is a word, simply because I am persuaded that it would not otherwise occur, still less recur, in this work. But this is a different matter.

⁷ As elsewhere, I am assuming here that the sentence under scrutiny consists of one or more morphemes whose status as such is not in question.

⁸ The great rarity of the collocation waste-paper-basket sentence must put us on our guard against a possible misconception. A very rare collocation may be perfectly clear in the appropriate context and may not involve us in any agonizing speculations about possible radical shifts of meaning of one or other of the words in it.

⁹ The only situation that I can think of (other than in a paper about language) where it would be wholly appropriate would be on the lips of someone in a pathological state of mind, e.g. in a delirium, where regularity of grammatical pattern and eccentricity of collocation often go together.

¹⁰ Another way of saying this is that if there were no restrictions on the collocability of molten (or any other word we may care to select), it would then have no meaning other than ‘grammatical’, i.e. what it had by virtue of whatever restrictions there still were as to the places in grammatical structure it was eligible to occupy. Cf. footnote 4.

¹¹ Or perhaps rather what Ludwig Wittgenstein calls ‘a complicated network of similarities overlapping and criss-crossing: sometimes overall similarities, sometimes similarities of detail.‘ See Philosophical investigations 1.66 (Oxford, 1953).

¹² It need not, incidentally, disturb us if some words seem to have more than one range, such as a dictionary would signalize by listing separate meanings in a succession of subheadings. This is precisely what we must expect, and we may often find it convenient to divide our range into subranges as circumstances and our intentions suggest.

¹³ Wittgenstein 1.67.

¹⁴ See the Oxford English dictionary s.v. feather sb. 16b.

¹⁵ There are of course pattern-extending tendencies also, which amounts to saying that grammar does not remain fixed either, but we are not concerned with these here; they form an ill-explored branch of the history of English. Cf. footnote 18.

¹⁶ It is in contexts like this that it is useful to keep in mind the distinction between the general and the special approach. One might say for example that sour lemon is a familiar enough collocation but that if the whole phrase is used in an unfamiliar context its impact may be considerable, e.g. if I use it in reference to my uncle. Strictly speaking, then, it seems that we should distinguish between familiarity from the point of view of the forms involved and from the point of view of the referent. In practice this is not important, because (unless ‘uncle’ figures only in the situational context and not in the text) at the dimension where we are dealing with uncle and sour lemon as themselves a collocation, we shall already have made a distinct step towards the unusual.

¹⁷ Examples such as these may well give us trouble the first time they are encountered; they then differ in this respect from the aged chemistry professor.

¹⁸ A matter quite as important as range-extending is the process whereby range is constricted through the abandoning of previously familiar uses, or the more complicated but also more common situation where the range is constricted at one point but extended at another. In all this there are parallels to what happens with grammatical patterns.

¹⁹ This of course applies also to lexical items as such and to grammatical patterns as well. Furthermore, in the case of lexis-range and pattern constriction, it may be the young who question the acceptability of something which is regarded as normal by the elderly.