Sometimes a morpheme changes form in different contexts. The different forms of a morpheme are called allomorphs. For example, the plural allomorph of English takes many different forms, as shown in (1). These include predictable variation, such as variation between voiceless /s/ after voiceless consonants as in cats in (1a), voiced /z/ after voiced consonants as in dogs in (1b), and epenthetic /əz/ after sibilants as in bushes in (1c). The English plural form also has unpredictable, or irregular, variation, including vowel changes as in foot versus feet in (1d), irregular suffixes such as ox versus oxen in (1e), and plural forms borrowed from other languages as with cactus versus cacti in (1f). Sometimes multiple strategies combine as with child versus children in (1g), which exhibits both an irregular suffix -ren and a vowel change from /aɪ/ to /ɪ/. Finally, sometimes the plural form is the same as the singular form, as with sheep. These are all examples of plural allomorphs in English.

We can classify allomorphs based on two properties. First, we consider the relationship between the different forms of the morpheme. If the different allomorphs are related to each other through a regular phonological pattern, they are called phonological allomorphs. If they are not, they are called suppletive allomorphs. Second, we consider the contexts that trigger allomorphy. Allomorphy can be triggered by phonological, morphological, or lexical conditioning.

Phonological allomorphs

When the form of an allomorph is predictable based on phonological context, it is called a phonological allomorph. In other words, phonological allomorphs are in complementary distribution based on a predictable phonological context. It is possible to write a morphophonological rule to describe where each allomorph is used.

For example, the plural allomorphs in (1a)-(1c) are phonological allomorphs. The plural morpheme surfaces as /əz/ after sibilants, as /s/ after other voiceless consonants, and as /z/ elsewhere. We can write two morphophonological rules which, taken together, describe this pattern, as in (2).

We can also write a distribution diagram, as in (3).

Suppletion

On the other hand, suppletive allomorphs are not predictable from phonological context.

Weak suppletion has some phonological similarity to the other allomorphs. Some example of weak suppletion can be identified in English past tense forms, shown in (4).

The present tense forms buy, catch, and teach have nothing in common with each other that would explain why they do not take the regular -ed past tense marker, which is why these past tense forms are considered suppletive and not phonological. However, each present tense and past tense pair share the same initial consonant—/b/ for buy and bought, /k/ for catch and caught, and /t/ for teach and taught—which is why these are considered weak suppletion.

In contrast, in strong suppletion, the allomorphs do not share any phonological similarity. For example, as shown in (5a), the root for the verb go and its past tense went have no phonological similarity. Likewise, shown in (5b), the root good has no phonological similarity with the comparative form better or the superlative form best. Therefore, good exhibits strong suppletion. On the other hand, the relationship between better and best could be considered weak suppletion, because they share the initial /bɛ/.

Conditioning

The choice of which allomorph is used where can depend on different factors: it can be phonologically, morphologically, or lexically conditioned.

When allomorphy is phonologically conditioned, it means the choice of allomorph depends on phonological context. For example, the regular forms of the plural allomorph are phonologically conditioned. The plural morpheme –s appears as [əz] after a sibilant, [s] after a voiceless consonant, and [z] elsewhere.

Morphological conditioning, on the other hand, depends on the morphological context, usually the grammatical properties of the morpheme, such as an inflectional category. For example, in Spanish, the verb meaning ‘to go’ uses the stem ir in the infinitive and future, va- in the present and imperfective past, and fu- in the perfective past. This is morphological conditioning because it depends on morphological features such as present, infinitive, or future.

Finally, lexical conditioning depends on the properties of the stem. For example, the Persian plural marker appears as –an for human nouns and –ha for non-human nouns, which is an example of lexical conditioning based on semantic properties. Lexical conditioning also includes any form that is not predictable (i.e., must be memorized). The irregular plural -ren in children is an example of lexical conditioning that is not predictable. There is nothing about the word child that predicts that it would have an irregular plural.

Check yourself!