Prosodic structure theory

Early work in generative phonology, for example, Chomsky and Halle (1968) and McCawley (1968), made it clear that in order to account for the manner in which rules of phonology apply in environments larger than the word, the sentence must be analyzed into a sequence of domains of different levels. In these works boundary symbols gave notational representation to domains, and there was no necessary coincidence between the boundaries of different domains. The theory that the representation of these phonological domains is instead a hierarchically arranged prosodic constituent structure, as proposed in Selkirk (1978, published in 1981) and developed elsewhere (see references), is based on a number of observations. A first observation is that the limits of higher order (larger) phonological domains systematically coincide with the limits of lower order domains. This leads naturally to a theory of the representation of these domains as a well-formed bracketing or tree, with each instance of a domain of a particular level being a constituent. A tree representation predicts this relation between domains, while boundary theory does not. Moreover, it seems to be the case that a higher order constituent Pn immediately dominates only constituents of the next level down Pn−1.