Defining and enhancing attributes for features

Abstract

We take the view that words are represented in memory as sequences of bundles of binary distinctive features. It is further assumed that each feature is defined by a quantal relation between an articulatory attribute and a corresponding acoustic property that provides acoustic cues to the feature. In the speaking process, other acoustic properties are introduced in addition to the defining acoustic attributes, and these contribute additional cues that enhance the perceptual saliency of the feature. For example, the defining attributes for the voicing feature is the presence or absence of glottal vibration during an obstruent consonant, but there are several possible enhancing properties, such as the presence of aspiration noise and a modified fundamental frequency in the following vowel. In running speech, the saliency of a feature may be in jeopardy, especially when there is interfering noise or when gestural overlap weakens some of the cues, particularly those derived from the defining gesture for the feature. An example is the casual production of batman where the closure for /t/ may be eliminated by gestural overlap, but two enhancing gestures (F2 transition for the alveolar and the glottalization for /t/) are preserved. [Work supported in part by NIH Grant DC00075.]