Magyar Fonetikai Füzetek 24. - Tamás Szende: Phonological Representation and Lenition Processes (1992)
1. Theoretical approaches to word-level phonological representation in post-SPE frameworks - 1.6. Dependency Phonology
tion predicts that the same structural properties recur at different levels. Structural properties postulated as unique to a particular level are unexpected and have to be supported by especially firm evidence of their unique appropriateness. Two levels of structure (syntax and phonology, for example) may be 'heteroplanar': although they share structural properties, the basic 'alphabets' of categories do not overlap (cf. Anderson—Durand 1986, 3—4). The crucial shared structural property is what is called the head—modifier (dependency) relation, e.g. the dominant constituent (the head) of a sentence is the predicate, modified by the subject; in turn, the head of the predicate is the verb, modified by e.g. the object, etc. The direction of modification may differ from level to level. Assuming the dominance principle and a unique alphabet, phonological representations will exhibit a novel kind of patterning. For the identification of segments, DP uses a small set of basic components that are meant to constitute a logically tight system. For instance, the components used to describe vowels are i_ (palatality or acuteness/sharpness), a_ (lowness or sonority), jj (roundness or gravity/flatness), cf. Anderson—Durand (1986, 23). (In the earlier Hungarian literature, Deme [1958] had proposed the same "phonologically utilized sound properties", with the additional property of 'duration'.) In DP, components are elementary units of phonological processes as well (the reduced form of English jto, [ta] , is derived from /tu/ by reduction or 'dearticulation', conceived of as a simple omission of the component u). The description of phonological forms involves two major principles in this framework, (i) Natural recurrence: phonological groupings (paradigmatic and syntagmatic) are not random: certain groupings recur; furthermore, phonological groupings (and the relationships between them) have a phonetic basis: they are natural (cf. Anderson 1980, 165). (ii) Natural appropriateness requires that a phonological notation should optimize the expression of such groupings (cf. Anderson—Durand 1986, 7). For instance: given that the feature [+high] characterizes a whole natural class of vowels, it is a recurrent property that has a consistent phonetic correlate, i.e. high tongue position, hence it is also natural. In addition, it shows clear affinity with some consonants that are also characterizable as [+high], both in paradigmatic (inventory-related) and syntagmatic (organizational pattern-related)