Definition

If the utterances of a language con­sisted merely of arrangements of phonemes, there would be no point in speaking or in listening. But people do speak and listen, and their oral communication transmits infor­mation and instructions and serves to coordinate their ac­tivities. That utterances can serve in this way is because they have another kind of structure in addition to the phonemic, one, a structure in terms of morphemes.

Morphemes are the smallest individually meaningful ele­ments in the utterances of a language.

To illustrate, we shall examine the following English sentence:

[³já n + ²trí jtsiz + owldə r + sí stə rz + vé rij + ²ná jsli²].

(John treats his older sisters very nicely.)

In order to determine the morphemes of which this sentence is composed, we pull out any portion and ask the following questions about it:

(I) Does the portion recur in various utterances, with approximately the same meaning? If the answer is no, then the portion we have chosen to examine is of no use to us, and we try another. If the answer is yes, then the portion is tentatively a grammatical form (or, for short, simply a form), but not necessarily a single morpheme. (It is unfor­tunate that we must include “tentatively” in the preceding statement, especially since the reasons for the reservation cannot be explained until § 19. In the meantime we shall proceed as though no reservation had been expressed.)

(II) Can the form be broken into smaller pieces, each of which recurs with approximately the same meaning, in such a way that the meaning of the whole form is related to the meanings of the smaller pieces? If the answer is yes, then the form is larger than a single morpheme (is a com­posite form) and we must subject each of the pieces, in turn, to the same two-step examination. But if the answer is no, then the form is itself a single morpheme.

Thus each portion we choose is shown, by Test I, to be either a bad choice or a grammatical form, and each grammatical form is shown, by Test II, to be either a com­posite form or a morpheme. By a series of such operations, we can discover all the morphemes of an utterance.

Let us apply the tests to the following extracts from our sample sentence: [já ], [já n + tr], [owldə r] and [sí stə r].

The first portion, [já ], fails Test I. It recurs, true enough — for example, in Jobs are scarce here, He’s a jolly old man, Two jars of shaving cream. But we detect no common feature of meaning in these utterances which could reasonably be assigned to the recurrent portion [já ].

The second portion also fails Test I. The portion recurs: John traded his watch for a pencil. If John tries that he’ll fail. From the broken demijohn trickled a stream of wine. But the requirement of similarity of meaning is not satis­fied.

Test I is quickly passed by [owldə r]. Its meaning in the original sentence is certainly much the same as in such sentences as He is older than I; The older of the two is a girl; I do declare, I’m getting older every day! In order to apply Test II, we must decide how to break [owldə r] up into smaller pieces. If we were working with an alien language we might have to test many alternatives — say [ow] and [ldə r], [owl] and [də r], and so on. Since we con­trol English natively we can avoid this complication and proceed immediately to the cut which we feel will yield positive results: [owld] and [ə r]. The former recurs, with reasonably constant meaning, in such sentences as He’s an old man, He’s the oldest of their three children, Jack is quite an active oldster. And the latter recurs in such sen­tences as When I was younger I enjoyed such things more, You should learn to enjoy the finer things of life. The evi­dence seems quite clear: older is more than one morpheme.

Similar testing of [owld] and [ə r] shows that each is only a single morpheme; older, then, is exactly two morphemes.

Finally [sí stə r]. This quickly passes Test I: My sister Eileen; OK, sister, get moving!; Sister Angela will be here in a moment. Turning to Test II, once again we have to decide what break-up to try. Let us first try sist- and -er, if only because this is much like the cut of older which proved fruitful.

Now there can be no doubt but that the string of pho­nemes [sí st] occurs in environments other than those in which it is immediately followed by [ə r], and it is equally obvious that the latter occurs where it is not preceded by [sí st]. Thus, for [sí st], we have He has a cyst which must be removed; I have a system, I can’t lose; Whipped cream consists largely of air; I don’t mean to insist. And for [ə r], in addition to the examples given earlier, we could find sentences involving brother, father, mother, daughter; ham­mer, butter, fetter, wither; singer, writer, actor, better.

But this is not enough. We get into trouble on the score of meaning, just as we did with the portions [já ] and [já n + tr] which we tested first. There seems to be no reasonable similarity of meaning between the sist- of sister and any of the other [sí st]’s illustrated. The words sister, brother, father, mother, daughter are all kinship terms, which means that they share some feature of meaning; on this basis one might want to extract the element -er as a morpheme carrying this shared feature of meaning. How­ever, to do so leaves us not only with a [sí st] which — in this meaning — seems not to recur, but also with similarly forlorn elements [brə ð ], [fað ], [mə ð ] and [dɔ t]. Thus it seems reasonable to conclude that sister should not be re­garded as a combination of smaller forms sist- and -er.

No other way of cutting sister into smaller pieces seems to have even the partial justification which we have found above for the cut into sist- and -er. We therefore decide to accept sister as a single morpheme.

Proceeding in this same way with all the different parts of our original sentence, we arrive finally at the following list of the constituent morphemes:

Note the following points:

First, the intonation must not be overlooked; we have taken it as a single separate morpheme.

Second, (5) and (9) are phonemically the same, but cer­tainly not the same morpheme, because of the difference in meaning.

Third, the breakdown of his [ı z] into hi- [ı ]and -s [z] may seem unconvincing. The [z] recurs, with exactlyme same meaning, in John’s book, the men’s room, and the like. But the [ı ] recurs only in him (as in hit ‘im).

If this evidence is enough to persuade us to break up (h)is, then maybe we want to break up very top, into a ver- which recurs in verity, veritable, perhaps veracious, and an element -y which recurs in pretty (pretty well) and perhaps elsewhere.

Marginal uncertainties of this sort are to be expected — in any language, not just in English. They must not be allowed to disturb us too much. Most problems of whether to cut or not are answered easily and quickly. Where there is conflict of evidence, it is often not very important which alternative we choose. The uncertainties lie in the nature of language, rather than in our method of attack.

Words

Idioms

A final lay assumption about “words”, which does not actually hold either for the layman’s words or for ours, is that they should always have some sort of meaning of their own, predictable in terms of their struc­ture if they are larger than morphemes, and reasonably constant from one occurrence to another.

Minimum free forms and lexemes also do not meet this requirement. The units which do are the least wordlike of any of the types we srtail discuss. The best approach to these units, which we shall call idioms, is via examples in some other language.

The Chinese form youtung has as ICs the two morphemes you ‘oil, grease’, and tung ‘large cylindrical container’; the first IС modifies the second, as black modifies cat in black cat or grease modifies rack in grease rack. Given this information, but knowing, nothing else about Chinese or the culture of China, we can venture a reasonable guess as to the meaning of youtung: ‘oil container’, ‘oil drum’, or the like. This guess is correct.

The ICs of Chinese mashang are ma ‘horse’ and shang ‘space on or above, top, ascend’. As in the preceding example, the first constituent modifies the second. Reasonable guesses at the meaning of mushang would be ‘horse’s back’, ‘horseback’, or possibly ‘on horseback’. These are wrong. The meaning is ‘quickly, right away’.

This meaning is not surprising when we remember that until recently the most rapid mode of travel was by horse. But it is one thing to consider a meaning reasonable after we know it, and quite different to deduce the meaning of a form from its structure. A native speaker of Chinese is no better off than we, for he can know ma and shang and still not understand mashang unless he has learned the meaning of the latter as a separate fact about his language.

Let us momentarily use the term “Y” for any grammat­ical form the meaning of which is not deducible from its structure. Any Y, in any occurrence in which it is not a constituent of a larger Y, is an idiom. A vast number of composite forms in any language are idioms. If we are to be consistent in our use of the definition, we are forced also to grant every morpheme idiomatic status, save when it is occurring as a constituent of a larger idiom, since a morpheme has no structure from which its meaning could be deduced.

Thus new is an idiom in She wants a new hat, but not I’m going to New York, because here it is part of the larger idiom New York. New York, in turn, is an idiom in the preceding sentence but not in The New York Times or The New Yorker, since in the latter expressions New York occurs as part of larger idioms. The advantage of this feature of our definition, and of the inclusion of mor­phemes as idioms when they are not parts of larger idioms, is that we can now assert that any utterance consists wholly of an integral number of idioms. Any composite form which is not itself idiomatic consists of smaller forms which are.

A composite form in another language cannot be called an idiom merely because its meaning seems queer to us.

The test must be applied within the language. French Elle est garde-malade ‘She is a nurse’ may seem peculiar to us because it contains no equivalent for English a, but this is the regular habit in French, and the sentence is no idiom. On the other hand, though French mariage de convenance finds its exact counterpart in English marriage of convenience, both the French and the English phrases are idioms.

An idiomatic composite form may coincide in morphem­ic shape with a form that is not idiomatic. White paper is an idiom when it refers to a certain sort of governmental document, but not when it refers merely to paper that is white.

A single form can be two or more idioms. Statue of Liberty is one idiom as the designation of an object in New York Bay; it is another in its reference to a certain play in football. Bear is presumably the same morpheme in wom­en bear children and in I can't bear the pain, but it is different idioms in these two environments.

Idioms are unwordlike especially in that they can be much larger than single words: Now is the time for all good men to come to the aid of the party. Yet some idioms are smaller than words. Bought, went, paid, sold, sang, rang consist of two morphemes each. One is, respectively, buy, go, pay, sell, sing, ring; the other, in all of them, is the “past tense” morpheme. In most occurrences, however, the meanings of the whole forms are predictable from the meanings of the constituents, so that the whole words are not idioms.

In theory, and largely in practice, idioms are the stuff of which dictionaries are made. The reason is obvious: a dictionary-maker need not include a non-idiomatic nonce-form, since a speaker of the language would never look up such a form. He would look up the component parts, if he needed to, and automatically know the meaning of the whole. In practice, of course, no dictionary is ever com­plete. There are far too many idioms in any language, and more come into existence every day.