Deriving Prosodic Structures

When we speak, we speak in prosodic chunks. That is, in the speech flow, we produce sound strings that are systematically parsed into intonational units. The parsing procedure not only eases the production of the speaker, but it also provides the hearer clues about how units of meaning interact in the articulated speech. The correspondence of the units of articulated speech (i.e. prosodic words, phonological phrases and intonational phrases) and the units of meaning (such as morpho-syntactic words, phrases and clauses / parentheticals) is a commonly observed phenomenon. Building on the observations on the alignment of units of articulated speech and units of meaning, I investigate the manner in which this correspondence is set and what aspects of the units of meaning affect the generation and the organisation of the units of the articulated speech. Challenging the previous assumptions of the theories that aim to predict a correspondence between morpho-syntactic words and prosodic words, I claim that such a correspondence is implausible, both theoretically and empirically. In the same vein, I conclude that not all properties of the units of meaning are reflected on the generation and organisation of sound units. For instance, not all clauses and parentheticals correspond to intonational phrases, and not all phrases correspond to phonological phrases. To account for the limited fashion of sound-meaning correspondence, I suggest a novel derivational account, in which the units of meaning are pruned unless they are relevant to the parser.