In Section 5.1 we defined the morpheme as the “smallest meaningful unit of language”, following a widely-used definition. This definition seems to work well in that it allows us to describe the internal structure of words in an insightful way. For example, we are able to describe a word like intolerable as consisting of the bound base toler-, which combines with the derivational suffix {-able} to form the derivative [[toler][able]], which in turn combines with the derivational prefix {in-} to form the derivative [in[[toler][able]]]. Likewise, we are able to describe an inflected word like dogs as consisting of a monomorphemic base (or: a root) dog, which combines with the inflectional affix {-s} ‘plural’ to form the word form [[dog]s].

This is correct as far as it goes, but it does not go quite far enough. Consider the following word forms:

They all contain the plural affix {-s}, but note that its form varies in a way that is obscured by the spelling: sometimes it is /s/, sometimes it is /z/ and sometimes it is /əz/. In other words, English has not one, but three forms that mean ‘plural’ — all three of them meet our definition of being a “smallest meaningful unit” of English. What is going on here? Are there three plural morphemes in English?

Not quite. If you look more closely, you will notice a pattern: /z/ is used whenever the base ends in a voiced phone (a vowel, as in tree or a voiced consonant, as in dog or lesson), /s/ is used whenever the base ends in a voiceless consonant other than /s/, /ʃ/ or /t͡ʃ/, and /əz/ is used whenever the base ends in /s/, /z/, /ʒ/, /ʃ/, /d͡ʒ/ or /t͡ʃ/. This should remind you of something — it is similar to the situation we saw in phonology, where different phones can represent a single abstract category called phoneme!

Just like the phoneme, the morpheme is best thought of as an abstract category containing one or more forms with the same meaning: /s/, /z/ and /əz/ are members of the plural morpheme, but none of them are the plural morpheme. We therefore need a more specific terminology to describe morphological structure, and we need a way to capture a situation like that of the English plural. Let us reserve the term morpheme for a category of forms that express exactly the same meaning, for example the category of forms expressing the plural meaning ‘more than one’. This morpheme contrasts in the same contexts with other morphemes, i.e. they are in contrastive distribution. For example, the plural morpheme contrasts with {-let} and {-ish} in books vs. booklet vs. bookish. Any specific meaningful unit of form can then be called morph, and morphs that belong to the same morpheme can be called allomorphs. These allomorphs are in complementary distribution, i.e., distributed in such a way that no two of them can occur in the same context and that we can predict which one of them will occur in any given context. In other words, /s/, /z/ and /əz/ are morphs, more specifically, they are allomorphs of the plural morpheme, and they occur in complementary distribution.

The distribution of allomorphs can be captured by allomorphic rules, using a notation broadly similar to that of allophonic rules. For the English plural, a first approximation could look as follows:

Inflectional morphemes are often represented in curly braces, often (but not always) with their function given in upper case, as we have done here. This convention does not work well for derivational morphemes — we will discuss them later. The allomorphs are often represented in square brackets (as we have done), but you will also often see them represented in slashes or without any additional notation instead.

The plural allomorphs in our rule are predictable on the basis of the phonetic context — such allomorphs are referred to as phonologically conditioned allomorphs. There is another type of allomorphy — consider the following examples:

Again, all of them contain the plural morpheme, but none of them behave according to the rule we have just posited. We could call them “irregular plurals”, as is often done, and not worry about them — there are exceptions to every rule. However in doing so we would ignore the fact that there is a certain systematicity behind these plurals: they are not the only cases with their respective ‘irregular’ plural morph. Other cases ending in /iː/ (like formulae) are, for example, antennae or larvae; another case ending in /aɪ/ (like stimuli) is cacti (from cactus); other cases ending in /ə/ (like bacteria) are criteria (from criterion) or referenda (from referendum), and other cases ending in /iz/ (like oases) are axes (from axis) or diagnoses (from diagnosis).

These cases are irregular in that you have to learn the individual words in which each of these plural forms occurs, but once you know this, they are regular in that not every word is different, but there is always a set of words sharing the same plural morph. Morphs like this are referred to as lexically conditioned allomorphs, as the context that determines which of them we find is the specific base itself. In this case, the allomorphic rule would look like this:

Note that in this solution, we assume that all bases are bound and that their singular forms are also inflected — we would have to have allomorphic rules for a {SINGULAR} morpheme, too. Such an analysis looks reasonable in the case of a base like formul-, stimul– or diagnos-, because these do occur with other affixes, such as {-ate} in formulate or stimulate or {-Ø} in diagnose, but most cases would be unique bases. Also, the singular is not usually marked in English, so such a rule would make things very complex. An alternative to such a description could be a more complex rule that first deletes the nucleus and coda of the final syllable of the singular, and then adds the respective plural suffix.

You might wonder whether it is really a good idea to treat these plurals as rule-based at all — clearly, all cases are loanwords from Latin, and the plural forms are Latin plurals. However, the fact that there are a large number of words with these plurals that were borrowed into English makes it plausible that speakers recognize this systematicity — hence, we should not treat the words as unanalyzable, and allomorphic rules seem to be a reasonable way to analyze them.

Even with plural forms that are not borrowed from Latin, we find cases that do not follow the general allomorphic rule introduced above. Consider the following cases:

The first case is special in that the plural is not marked by an affix at all. We can deal with this by positing an ‘invisible’ zero allomorph (remember the discussion about conversion in Section 5.6):

With children, there are two issues to consider. One is the plural allomorph /ən/, which we also find in brethren (from brother) and sistren (from sister) and oxen (from ox):

The second issue is the stem, which also changes in some words — /tʃaɪld/ child becomes /tʃɪl.dr/, /ˈbɹʌð.ɚ/ brother becomes /ˈbɹeð.ɹ/, and /ˈsɪs.tɚ/ sister becomes /ˈsɪ.stɹ/. These are also cases of allomorphy, concerning the root rather than an affix. They can be dealt with in the same way:

The case of leaves is different again: here, we also have root allomorphy, but the plural suffix itself is regular once we take the allomorphy into account. The word leaf ends in a voiceless fricative, so we would expect the plural allomorph /s/, giving us /liːfs/ (as in /ɹiːfs/ for reefs). Instead, the root changes to /’liːv/, so we get the plural allomorph /z/, as expected after a voiceless phone. Other words that behave similarly are dwarf/dwarves, half/halves, hoof/hooves, knife/knives, loaf/loaves, self/selves, thief/thieves, wife/wives, wharf/wharves, wolf/wolves. Again, we can capture the allomorphy of the root in a rule, e.g.

The rule has to be posited for each individual word rather than, for example, all roots ending in /f/, because it only applies to a subset of such roots — the words bluff, chef, chief, cliff, cough, cuff, nymph, staff, photograph, proof, riff, and roof, for example, do not change and have the expected allomorph /s/.

Finally, the word mice is a true exception. There are a few other words where the plural is indicated by a change in the vowel of the root, e.g. feet, geese, mice, men, women and teeth. However, there is no systematic relationship between the vowel in the singular and the plural, so we would need a rule for every word pair. It is more plausible to treat the plural forms as words whose meaning incorporates that of the singular root and the plural affix, e.g. [maɪs] ‘mouse + PLURAL’. Such cases are sometimes referred to as stem mutation (stem is a special term that some linguists use for a base of an inflectional affix) or internal modification.

That a single, unanalyzable morph represents two morphemes simultaneously is not unique to these stems. It is also the case, for example, with the morph {-s} as a marker of third person in the present tense (it represents the morphemes {3rd PERSON} {SINGULAR} and {PRESENT} simultaneously).

An extreme case of allomorphy is suppletion, where the root is not only modified but replaced by another one in a specific morphological context. For example, the word person has two plurals: one with the allomorph /z/ — persons —, and one where the root is replaced — people ‘person + PLURAL’. Suppletion is rare, but there are some verb roots in English that form different tenses by suppletion — e.g. go, with the past tense form went ‘go + PAST’, or be, with, among others, is ‘be + 3rd PERSON + SINGULAR + PRESENT’.

Let us now look at a case of a derivational morpheme with several allomorphs. Consider the following data:

All of these words contain the prefix {in-} that we already saw in Section 5.3. However, its form varies — in (4a), (4b) and (4c), it is /ɪn/, in (4d) in is /ɪm/, and in (4e) to (4h) it is simply /ɪ/. If you look at the phonological context, you will see that we can predict the form of the affix: before nasals and approximants, it is /ɪ/, before [p], it is /ɪm/, and everywhere else, it is /ɪn/:

Note that we do not simply represent the morpheme by its meaning, as we did in the case of the inflectional affix {PLURAL}. The reason is that meanings of derivational affixes are often difficult to capture — we could represent the affix {IN-} as {NEGATIVE}, but then we could not distinguish it from {non-}, {dis-} and other morphemes with very similar meanings. Thus, it is better to choose one of the allomorphs and use it as the name of the morpheme — in uppercase letters, if you want to be very clear that you are talking about the morpheme, not a specific morph.

← Previous section Next section →

CC-BY-NC-SA 4.0, Written by Anatol Stefanowitsch