Allomorphs are variant forms of a single morpheme, triggered by phonological context or morphological properties. The English plural -s surfaces as [z] after voiced sounds, [s] after voiceless sounds, [əz] after sibilants—this phonologically-conditioned allomorphy follows from phonological rules. Suppletive allomorphy (go/went) and morphologically-conditioned allomorphy (English man/men) require lexical specification. Understanding allomorphy determines the inventory of underlying morphemes.
Identify allomorphic variants of a morpheme across contexts, determine whether conditioning is phonological, morphological, or arbitrary, and write rules or constraints deriving the distribution.
From your study of morpheme types and morphological structure, you know that words are built from smaller meaningful units — morphemes — and that these units combine in systematic ways. Allomorphy is what happens when the same morpheme shows up in different phonological shapes depending on context. The key conceptual move is distinguishing the morpheme (the abstract unit of meaning) from the morph (the actual phonological form). A single morpheme can have multiple morphs — those are its allomorphs.
The English plural is the clearest example to start with. The plural morpheme is one thing — it means "more than one" — but it surfaces as three different sounds: [z] in *dogs*, [s] in *cats*, and [əz] in *buses*. This is phonologically-conditioned allomorphy: the choice of allomorph is fully predictable from the phonological environment. After a voiced consonant or vowel, you get [z]; after a voiceless consonant (other than a sibilant), you get [s]; after a sibilant, you get [əz]. No one taught you this rule explicitly — you generalized it from input as a child — but it operates consistently. The analysis task is to take the three surface forms, identify the conditioning environment for each, and reduce them to a single underlying representation plus a rule (or constraint set) that derives them.
Suppletive allomorphy is the other end of the spectrum. *Go* and *went* are allomorphs of the same verb — the past tense of "go" — but no phonological rule derives *went* from *go*. The relationship is historically arbitrary, a relic of the merger of two different Old English verbs. These must be lexically specified: the grammar simply stores the pair. Similarly, *man/men* and *mouse/mice* are morphologically-conditioned allomorphs of the noun plural, where the alternation follows from the word's lexical class (historically, the old Germanic *umlaut* plural), not from the phonological environment. The analytical challenge is figuring out *which* type of conditioning is at work in any given case.
Why does this matter? Because identifying the correct allomorphic analysis determines your morpheme inventory — the set of underlying units your grammar operates over. If you treat [z], [s], and [əz] as three separate plural morphemes, your grammar becomes redundant and misses the generalization. If you posit one underlying form and a phonological rule, you capture the generalization with less machinery. This is the core principle of morphophonological analysis: prefer the analysis that captures alternations as regular rules over one that requires arbitrary lexical listings. Suppletive forms are the residue that cannot be derived by rule — they are the cases where the historical generalization has been lost, and the grammar has no choice but to memorize.