Is every morpheme pronounced the same in all contexts? If it were, most phonology texts could be considerably shorter than they are! In fact, many morphemes have two or more different pronunciations, called allomorphs, the choice between them being determined by the context. These include some of the commonest morphemes in the language, as I will illustrate directly. I will then discuss in more detail what aspects of the context can influence the choice of allomorph.

How are the plurals of most English nouns formed? If one compares cats, dogs and horses with cat, dog and horse respectively, the obvious answer is: ‘by adding -s’. But English spelling is notoriously unreliable as a guide to pronunciation. In fact, this -s suffix has three allomorphs: [s] (as in cats or lamps), [z] (as in dogs or days), and [Iz] or [əz] (as in horses or judges). Is it, then, that everyone learning English, whether natively or as a second language, must learn individually for each noun which of the three allomorphs is used in its plural form? That would seem extremely laborious. In fact, it is easy to show that the three allomorphs are distributed in an entirely regular fashion, based on the sound immediately preceding the suffix, thus:

• when the preceding sound is a sibilant (the kind of ‘hissing’ or ‘hushing’ sound heard at the end of horse, rose, bush, church and judge), the [Iz] allomorph occurs

• otherwise, when the preceding sound is voiceless, i.e. produced with no vibration of the vocal folds in the larynx (as in cat, rock, cup or cliff ), the [s] allomorph occurs

• otherwise (i.e. after a vowel or a voiced consonant, as in dog or day), the [z] allomorph occurs.

In effect, without realising it, we pay attention to these phonological characteristics of the noun when deciding which allomorph to use – though ‘decide’ is hardly the right word here, because our ‘decision’ is quite unconscious. Another very common suffix with phonologically determined allomorphs is the one spelled -ed, used in the past tense form of most verbs. Its allomorphs are [t], [d] and [Id] or [əd]; determining their distribution is left as an exercise, whose solution is provided later.

One may be tempted to think that the allomorphy involved here (i.e. the choice of allomorphs), because it depends so much on phonology, is not really a morphological matter at all. But that is not quite correct. Consider the noun lie meaning ‘untruth’. Its plural form is lies, with [z] – just as predicted, given that lie ends in a vowel sound. But this is not because either [s] or [əz] would be unpronounceable here, or would break some rule of English phonology. If we experiment by replacing the [z] of lies with [s], we get an actual word (lice, the plural of louse), and replacing it with [əz] we get what is at least a possible word (it might be the plural of an imaginary noun ‘lia’) – and is an actual word (liars) in those dialects of English where liar is pronounced without an r-sound. So phonologically determined allomorphy need not just be a matter of avoiding what is phonologically prohibited.

It is not only phonology that may influence the choice of allomorphs of a morpheme. Instances where grammar or vocabulary play a part in the choice are extremely numerous in English. We will do no more than skim the surface of this huge topic. We will look first at a set of examples that involve both grammar and vocabulary, before showing how a morpheme’s peculiar allomorphy can be crucial in establishing its existence.

The words laugh and cliff both end in the same voiceless consonant (despite what the spelling may suggest!). Therefore, according to the formula given above, the allomorph of the plural suffix that appears on them should be [s]. And this is correct. But what about wife and loaf ? These end in the same voiceless consonant as laugh and cliff; yet their plurals are not *wifes and *loafs but wives and loaves. (The asterisk is a conventional symbol to indicate that a linguistic expression (a word, phrase or sentence) is unacceptable for some reason to do with grammar or with the structure of the language generally, rather than for reasons such as truthfulness or politeness.) In fact, there are quite a few nouns which, in the singular, end in a voiceless f, s or th sound but which change this in the plural to the voiced counterpart (not always reflected in the spelling). Nouns that behave like this in most varieties of English are knives, lives, hooves, houses, paths and baths. However, there are also exceptions to this ‘rule’: apart from laugh and cliff, already mentioned, one can think of fife and oaf, which both form their plural with [s]. What’s more, wife, knife and the rest do not use their voiced allomorph (wive-etc.) before any morpheme except plural -s – not even before the ‘apostrophe s’ morpheme that indicates possession, as in my wife’s job. So the allomorphy here is determined both lexically (it is restricted to certain nouns only) and grammatically (it occurs before the plural suffix -s but not before other morphemes). This state of affairs suggests a refinement to the bound-free distinction: as a morpheme, wife is clearly free, but, of its two allomorphs wife (with final [f ]) and wive (with final [v]), only the former is free, while the latter is bound.