Prosodic complexity and processing complexity: evidence from language impairment

JOHN HARRIS, NICHOLA GALLON & HEATHER VAN DER LELY

In segmental phonology, there is a long tradition of linking markedness with representational complexity (for example, as reflected in the number of feature specifications a segment bears). The marked nature of more complex representational entities is often assumed to correlate with some notion of functional complexity or difficulty, placing an increased burden on articulatory effort, auditory perception or phonological processing. The link between complexity and markedness does not obviously carry over into prosodic structure. For example, while a binary branching syllable onset is both more marked and more representationally complex than a non-branching onset, a binary branching foot is less marked than a degenerate foot.

Unmarked binarity can plausibly be linked to constraints on the canonical prosodic shapes of different types of morpheme. One of these constraints requires lexical heads to branch, with the result that words are minimally bimoraic or bisyllabic. In many languages, this minimal structure coincides with the stress foot. What contributes to metrical complexity is any structure that augments this minimal shape, for example through the adjunction of unfooted syllables to word edges.

The paper reports the results of an English non-word repetition experiment designed to illuminate the influence of prosodic complexity on phonological processing. The subjects fall into three groups: one showing evidence of specific language impairment (SLI) and two age-matched control groups of typically developing children. Non-word stimuli were systematically varied along three syllabic and two metrical parameters, each representing a binary opposition between an unmarked and a marked structure: branching vs. non-branching onset; open vs. closed rime; word-final V vs. C; presence vs. absence of a left-adjoined unfooted syllable; presence vs. absence of a right-adjoined unfooted syllable.

The main results can be summarised as follows. Unlike the two control groups, the SLI group showed an incremental decrease in the number of correct responses as the number of marked prosodic structures per non-word increased. Increasing metrical complexity had a greater negative effect on overall performance than increasing syllabic complexity, particularly for the SLI group. While some of the errors triggered by metrical complexity affected metrical structure itself (in the form of weak-syllable omission), the bulk involved inaccurate renditions of the syllabic and/or segmental content of the relevant stimuli. The results support the conclusion that prosodic complexity can affect non-word repetition accuracy independently of string length.

To appear in Revista da Associação Brasileira de Lingüística

Back to John H publications list