John Harris (johnh@linguistics.ucl.ac.uk)
University College London
Research has largely vindicated Jakobson's claim that explicit parallels can be drawn between the phonological preferences exhibited by primary grammars and the sorts of phenomena encountered in language impairment and the early stages of language acquisition. The congruities are particularly striking in the realm of syllable structure. Broadly speaking, any syllabic configuration that is cross-linguistically disfavoured tends to be suppressed in emergent or disordered approximations of primary grammars that possess it. The best known examples involve the syllabification of consonants: the relatively low incidence of consonant clusters and word-final consonants in the world's languages is matched by the relatively high incidence of consonant deletion and consonant-supporting vowel epenthesis in language acquisition and impairment. These facts confirm the unmarked status of simplex onsets, open syllables and vowel-final words.
The general preference for non-branching constituent structure that is suggested by the marked nature of complex onsets and closed syllables extends to nuclei: only a minority of the world's languages possess a contrast between short vowels (non-branching nuclei) and long (branching). If the Jakobsonian insight is of general validity, we should expect vowel-length distinctions to come under pressure to restructure in language acquisition and disorder. Although this matter has not received much attention in the relevant literature, there is evidence to suggest that the expectation is borne out.
Part of this evidence is presented here in the form of two case studies of children who were initially diagnosed as presenting with 'vowel' impairments. Both disorders can be shown to have a syllabic basis, stemming specifically from a failure to achieve branching nuclear structure. Left to its own devices, this prosodic deficit might have been expected to produce across-the-board vowel shortening. In both cases, however, this result is apparently forestalled by an independent requirement not to fall below a minimum word size of two morae (interpreted here as a bi-nuclear licensing domain).
In one case, adult long vowels and diphthongs are rendered as two short vowels separated by a glottal stop, i.e. as two non-branching nuclei flanking an onset with minimal melodic content. In the other, the second portion of English up-gliding vowels is hardened to a homorganic stop. In this instance, it can be argued, the bimoraicity of the adult branching nucleus is retained in the child's system as a non-branching nucleus followed by a non-nuclear position.
Each of these systems can be described as parochially deviant; that is, it belongs to the set of possible primary grammars but happens not to coincide with the ambient adult grammar. Both take a marked syllabic option (branching nuclei ON) and restructure phonological output in terms of an unmarked option (branching nuclei OFF) that has abundant precedents amongst the world's primary systems.
Besides providing external support for the independence of syllabic structure, the case studies confirm the validity of prosodic models which recognise the nucleus as a distinct category. The studies also raise the perennial question of why certain melodic properties (in this case stopness) are restricted to certain syllabic positions.