Enhancement and Overlap in the Speech Chain

A model of speech production is proposed in which the input is a planning stage at which lexical items are arrayed, accompanied by the full panoply of phonological representations from distinctive features to their attendant tree structures. A set of instructions for control of the vocal tract is calculated leading to a sound output. Two parallel processes are involved in the calculation of these instructions, both of which replace the planning-stage representation by the appropriate motoric instructions. One of these processes is universal and involves replacing each distinctive feature with an appropriate set of motoric instructions, either unmodified or modified by the process of overlap. We postulate a parallel language-specific process that is sensitive to those features in danger of losing their perceptual saliency as a consequence of the environment in which they appear. This process, referred to as enhancement, adds additional motoric instructions to enhance the saliency of the jeopardized features. We provide a number of examples to illustrate how enhancement works. We conclude from these examples that whereas defining gestures related to distinctive features are, in many instances, weakened or even absented from the speech stream, enhancement gestures, once added to the set of motoric instructions, appear never to be subject to obliteration by overlap.