In linguistics, “prosody” refers to sound patterns in language involving more than a single segment or phoneme. Since the early 1980s, the study of prosody has blossomed, both in linguistics and in the allied areas of computer speech analysis and synthesis, adult sentence processing, infant speech perception, and language production. The study of prosody has provided insight into the word and sentence productions of young children with normally developing language and language disorders. In particular, prosody has proven a useful tool for examining children's “deviant” utterances; that is, utterances that deviate from what we would expect from an adult speaker with normal speech and language. For example, a child's production of “blue” as “bu” will be considered a deviant utterance for purposes of this discussion.

The article begins with a brief overview of two aspects of prosody, syllable shape and meter, that have been the focus of many studies of child language production. It then provides examples of some recent studies that demonstrate effects of syllable shape and meter on deviant productions of children with normal and disordered language. To foreshadow the general finding across the studies presented here, we state that syllable shapes and metrical patterns that are frequent across the world's languages or in the language the child is learning are most resistant to deviations.

Beginning with syllable shape, it has been noted that all languages of the world have syllables comprising a consonant plus a vowel (CV). Only some languages allow additional syllable shapes, such as V, VC, CVC, CCVC, and so on. In linguistic terms, the CV syllable shape is said to be “unmarked.” Even when languages allow syllable shapes other than CV, the shapes are often restricted. For example, Japanese allows only CVC syllables ending in /n/.

Furthermore, in languages like English that allow consonant clusters (syllables of the shape CCVC, CVCC, etc.), these clusters generally conform to a sonority sequencing principle (e.g., Hooper, 1976). Briefly, sonority may be conceptualized as the openness of the vocal tract for a particular segment. Vowels are the most sonorous; glides are less sonorous, followed by liquids, nasals, fricatives, and stops. Clements (1990) argues that an ideal syllable structure is one in which a sequence of segments increases from the onset to the vowel with no or a minimal decline from the vowel to the coda. English word-initial consonant clusters such as /pr/ and /kw/, which comprise a stop plus liquid or glide, are consistent with this principle, as are English word-final clusters such as /rp/ and /nt/.

The foregoing discussion of syllable shapes concerns what is allowed in words of a language. It is also important to note that even languages that allow a variety of syllable shapes nevertheless have strong statistical tendencies toward particular shapes. For example, English CVC words are very likely to end in /t/ and very unlikely to end in /d/. These statistical tendencies are the subject of a growing interest in researchers studying prosody and its role in child language production.

Let us now turn to meter. In many languages, multisyllabic words exhibit a characteristic stress pattern. For example, the majority of words in English have the pattern found in “apple” and “yellow,” that is, a stressed or strong syllable followed by an unstressed or weak syllable. Stressed syllables are louder, longer, and higher in pitch than unstressed syllables. The frequency of the strong-weak word pattern is consistent with the observation that the basic unit of stress in English is a trochaic (strong-weak) foot, with a foot defined as a grouping of a single strong syllable plus adjacent weak syllables.

Feet not only explain the dominant stress pattern of words in a language, they also help us to understand how lexical words like nouns and verbs combine with grammatical words like determiners and auxiliary verbs in phrases. For example, when we say phrases like “drink of water” and “pick a card,” we tend to combine the grammatical words (in this case “of” and “a”) with the preceding strong syllable, even though these words belong syntactically with the following word (in a prepositional phrase and noun phrase, respectively). That is, English speakers tend to create trochaic feet whenever they can, giving the language a characteristic metrical pattern.

The discussion up to this point has revealed that languages of the world and particular languages are biased toward specific syllable shapes (e.g., CV) and meters (e.g., trochaic). Beginning with syllable shape, let us now consider how these prosodic biases affect the productions of children with normally developing and disordered language. Two of the most frequent syllable shape deviations from a standard target produced by children are final consonant or coda deletion and consonant cluster reduction. With respect to coda deletion, this phenomenon has been viewed as one in which the speaker is resorting to the most common syllable shape, CV. However, recent studies indicate that there are significant differences in the rate at which children omit different codas in different prosodic environments. Zamuner and Gerken (1998) reported that normally developing 2-year-olds produced more codas and more coda types on nonsense words when the coda occurred in a stressed syllable (either on a monosyllabic item or an item with a weak-strong stress pattern, e.g., //). Zamuner (2001) discovered that children from the same population produced obstruent codas, which are more frequent in English, sooner than sonorant codas on CVC nonsense words. She also found that the same coda was produced less frequently when it occurred in nonsense names exhibiting less frequent biphones (e.g., CV, VC) than more frequent biphones.

With respect to consonant cluster reduction, several studies have shown a role for syllable shape, and in particular sonority sequencing, in this phenomenon (e.g., Barlow and Dinnsen, 1998; Ohala, 1999). These studies have revealed that children with normal language and language disorders were more likely to produce the least sonorous consonant of an initial cluster and the least sonorous consonant of a final cluster. That is, they produced CV sequences that were closer to the ideal syllable shape suggested by Clements (1990).

Turning now to the role of meter in language production, several researchers have noted that children are more likely to omit weak syllables from the beginning of words like “giraffe” and “banana” and, more generally weak syllables that do not belong to a trochaic foot (e.g., Wijnen, Krikhaar, and Den Os, 1994). The bias to produce trochaic feet has also been observed at the level of sentence production, where the determiner “the” is more likely to be preserved in a sentence like “He pats the zebra” (“pats the” forms a trochaic foot) than “He brushes the bear” (the syllabic verb inflection makes the formation of a trochaic foot containing “the” impossible; Gerken, 1996). Thus, the effects of prosody are not restricted to what have traditionally been considered phonological deviations but extend to morphosyntactic deviations as well. It is interesting to note that not all languages show as strong a bias toward trochaic feet as English does. For example, Spanish has many words like “banana,” which exhibit a weak-strong-weak pattern. Spanish-learning children have been shown to produce determiners at an earlier age than their English-learning counterparts, again suggesting a role for prosody in children's morphosyntactic development (Lleó and Demuth, 1999).

At least some of children's weak syllable omissions appear to occur during late stages of language production rather than during utterance planning, as evidenced by work by Carter (1999). Normally developing 2-year-olds and older children with language impairment produced sentences like “He kissed Cassandra” and “He kissed Sandy.” Note that the former sentence type was frequently produced with the first syllable of the name omitted (Cassandra → Sandra). Acoustic measurements revealed that, even though the two types of sentences contained the same number of overtly produced syllables (four), children produced the first sentence type with a longer duration, suggesting that they reserved a timing slot for the syllable they eventually omitted. One possible source of weak syllable omissions is a lack of complete control over the motor sequences involved in producing trochaic vs. weak-strong feet (Goffman, 1999).

Finally, several studies have revealed joint effects of syllable shape and meter on deviant utterances. In the Zamuner and Gerken study discussed above, children showed different rates of coda preservation for strong and weak syllables. Ohala (1998) found that young children with normal language were less likely to reduce word-medial consonant clusters in words with a strong-strong stress pattern. In a study of weak syllable omission in young children with normal language, Kehoe and Stoel-Gammon (1997) noted more omissions of the middle syllable of words like “elephant,” which exhibit a strong-weak-weak pattern, if the syllable began with a sonorant consonant. Carter (1999) found that adults with a variety of types of aphasia were more likely to omit word-initial weak syllables with V and VC syllable shapes than CV shapes. It seems likely that such results would be found in children with normal and disordered language as well.

In summary, linguistic studies of canonical syllable shapes and metrical patterns across languages and within particular languages provide the tools for fine-grained analyses of deviant forms produced by children with normal and disordered language. The results of these analyses feeds a growing consensus that those forms that are very frequent in languages of the world or in the child's target language are generally more robust and less susceptible to deviations from the accepted standard. Further research is needed to reveal the mechanism underlying prosody's clear effect on language production (see also prosodic deficits).