In this article I discuss some aspects of a model of segmental structure, in particular the representation of vowels. The central claim is that vowels can be represented in terms of three unary primitives, organised in a binary structure involving both DOMINANCE and DEPENDENCY relations.

My goal is to explore the consequences of this claim. Certain aspects of the proposal differ in matters of principle from the theory of segmental structure advanced in Chomsky & Halle (1968; henceforth SPE), and subsequent work in this tradition up to and including the ‘geometrical model’ as developed in Steriade (1982), Clements (1985), Sagey (1986), McCarthy (1988) and Halle (1988), even though it is argued in some of these works that certain features are unary, that they are organised hierarchically and that certain features enter into dependency relations.

In this essay I suggest that the conventional bifurcation of geometrically organised feature bundles into ‘Laryngeal’ and ‘Supralaryngeal’ categories, advocated most prominently in the provocative work of Clements (1985, 1987), Sagey (1986) and Archangeli & Pulleyblank (forthcoming), should be modified to eliminate the Supralaryngeal element altogether. Specifically, the proposal is that the category Place and the features of manner be removed from subordination under the Supralaryngeal node, as in the conventional configuration displayed in (1), and instead be subsumed directly under Root, as depicted in (2):1

While the use of categorical features seems to be the appropriate way to express sound patterns within languages, these features do not seem adequate to describe the sounds actually produced by speakers. Examination of the speech signal fails to reveal objective, discrete phonological segments. Similarly, segments are not directly observable in the flow of articulatory movements, and vary slightly according to an individual speaker's articulatory strategies. Because of the lack of a reliable relationship between segments and speech sounds, a plausible transition from feature representation to the actual acoustic signal has proven elusive. This paper utilises a theory of information processing, known as PARALLEL DISTRIBUTED PROCESSING (PDP) NETWORKS (also called neural networks), to propose a model which begins to express this transition: translating the feature bundles indicated in a broad phonetic transcription into continuous, potentially variable articulator behaviour.

This article is concerned with the internal structure of coronals. Although coronals are often considered the least marked consonants on the basis of frequency and acquisition, current models of feature geometry assign coronality no special status among places of articulation.1 In this paper, we argue that the structure of coronals differs from that of other consanats in that coronals lack a Place node altogether.

Recent work by Clements (1985), Sagey (1986), Steriade (1987a) and others has shown clearly that distinctive features are hierarchically organised, and that the hierarchy includes a Place of Articulation constituent. Proposals differ, however, as to the organisation below this Place node. Clements (1985) suggests that there is a Secondary Place node dominating the vowel features, but that [anterior], [coronal] and [distributed] are directly dominated by the Place node itself. Sagey (1986) has argued that there are distinct Articulator nodes, Labial, Coronal and Dorsal, each of which dominates certain binary features, respectively [round]; [anterior] and [distributed]; and [high], [back] and [low]. Dorsal is thus present for both velar consonants and vowels. Steriade (1987a) modifies the Sagey model by adding a Velar node for velar consonants, distinct from the Dorsal node for vowels.