885095974

speech production. Further, organizational principles or neural control strategies for speech generated from kinematic observation without concomitant electromyographic observation may result in unrealistic theo-rctical abstractions. The utility of including, in a generał sense, biomechanical and electromyographic levels of analysis in the development of detailed theories and models of speech production will be presented. Tentative speculations on characteristics of the speech motor control process will be discussed. [Work supported by NIH.]

9:20

2SP4. On the analysis of speech mo?ements. John R. Westbury (Waisman Ctr., Univ. of Wisconsin, 1500 Highland Ave., Madison, Wl 53705-2280)

New technological developments have madę it possible to compile, in one aftemoon, a speech kinematic data set tliat once required 3 months of tedious, exhausting effort. In a wcek’s time, sets that once required a year, and in a month, sets that once rcquired several years, can be compiled. To those old enough to have done things in the old ways, this is surcly remarkable, but perliaps also unsettling because there are few sound traditions to guide the analysis of this rapidly expanding articulatory database. Hcretofore, a central feature of many analysis schemes has been a heavy reliance on identifying “magie*’ moments in time and places in space—easily defined “events” in the kinematic signal streams. Articulator positions and velocitics associated with these events, and time and distance intervals between them, then provided a parsimonious characterization of the movements. This analytic approach is certainly useful, but also troubling, first because the functiona) significance of the measurement events is unknown; second, because their locations are sensitive to the essentially arbitrary conventions that must first be established to represent the data; and finally, because the few discrete measures that are madę may under-represent the phenomena of interest, and lead to flawed generalizations. Examples will be providcd to illustrate these weaknesses of the classical approach, and partia! rcmedies will be suggested. It is high time that we talk about methods.

9:45

2SP5. The degrees of freedom in controlling articulations. Keith Johnson, Peter Ladefoged, and Mona Lindau (Dept. of Linguistics, UCLA, Los Angeles, CA 90024-1543)

What do speakers control in the production of vowels? Data from two groups of five speakers saying words containing ten diflferent American English vowels indicate that in saying the same vowcl there is a great deal of variation between speakers in the articulatory gestures used, but considerablc consistency within speakers in the articulations used. Ali ten speakers had constrictions in similar regions of the vocal tract for each vowel; but the relation btween tongue movement and jaw movement was variable. Principal components analyses of the locations of six pellcts showed that. for one group, two of the five subjecLs used the jaw and tongue in combination, and their vocal tract shapcs could be described in terms of only two components. For the other three subjects, three components were needed to account for the same amount of variance, with the third component distinguishing tense and lax vowels. The other goup of five subjects was analyzed separately. Two subjects showed good correlation between tongue height and jaw height; for the other three, tongue height was not well predicted from jaw height, except in the case of back vowels. These speakers used various combinations of articulatory mechanisms for separating tense and lax vowels. Lip positions were also examined. Even in the case of back vowels, only some of the speakers had good correlation between lip opening and protrusion. In conclusion, the evidence that vowel production is goal oriented is discussed using targets learned as auditory goals but produced as individually controlled articulatory gestures. [Work supported by NIH.]

10:10

2SP6. The control and kinematics of human jaw movement. David J. Ostry, J. Randall Flanagan, and Anatol G. Feldman (McGill Univ., Montreal H3A 1B1, Canada)

The kinematics of two-dimensonal human jaw motion are presented based on x-ray microbeam record-ings. The relationship between jaw translation and rotation is described and cxperimental records are compared with simulations based on the equilibrium point hypothesis (A model). In generał, jaw rotation and translation were found to start and end simultaneously and straight linę paths were observed when rotation was plotted agaiust translation. Several manipulations suggest that jaw rotation and translation are separately controlled. For example, when jaw movements in speech were examined, the slope of the relationship between rotation and translation varied with the consonant but did not depend on the vowel or speech ratę. The kinematic pattems of jaw motion are well accounted for by the A model. The model demonstrates that separate central commands can be defined associated with jaw translation, jaw rotation, and coactivation of muscles without motion. Central commands may be superimposed to produce combinations of rotation, translation. and coactivation. Empirical patterns can be captured by the model under the assumption of simple constant vclocity shifts in equilibrium govemed by central commands.

1870

J. Acoust. Soc. Am., Voł. 09, No. 4, Pt 2, April 1991

121st Meeting: Acoustical Sodety of America

1870