Gestural aggregation in speech

Abstract

It is well known that the units of speech are not produced strictly sequentially but overlap with each other; this phenomenon has been referred to as coarticulation, coproduction, or blending. Thus the vocal tract shape at any one time during speech represents an aggregate of gestures associated with different production units. The present experiment examines one intra-articulator example of temporal overlap, namely, the combination of successive opening and closing gestures in the larynx. Two subjects produced the utterance “Kiss Ted” at several different speaking rates and with stress on the first or second word. Laryngeal abduction-adduction was monitored using transillumination and fiberoptic video recording. At slow rates, two separate opening gestures occur associated with the fricative and the stop, respectively. At fast rates, a single smooth gesture is seen. At intermediate rates, partially overlapping gestures are found. These findings suggest that the two underlying gestures are being blended into a single aggregate and that the aggregation operation may be some type of summation. It is hypothesized that blending of overlapping gestures is a general phenomenon in speech production that accounts for commonly observed types of coarticulation. [Work supported by NINCDS Grants NS-13617 and NS-13870 to Haskins Laboratories.]