Testing Theories of Speech Production: Implications of Some Detailed Analyses of Variable Articulatory Data

Abstract

Because of indirect relationships between theories and data in speech production, most theories are difficult to test. As an example, a general theoretical overview is outlined which suggests that speech production consists of sequences of movement ensembles from one set of articulatory goals to the next. A number of factors influence the nature of the goals; some of the most important are quantal or non-linear relationships among properties of the production mechanism, the acoustic characteristics of the resulting sounds and capabilities of the perception mechanism. The programming of articulatory movements is a function of: the nature and sequencing of articulatory goals, the moment-to-moment state of the vocal tract and the particular acoustic requirements of individual utterances. Three examples of articulatory data are presented, each of which bears in some way on these ideas and also contains a great deal of variability. One set of data suggests a mechanism for the definition of articulatory goals; the second suggests possible influences on the strategy of anticipatory coarticulation; and the third implies that variability in articulatory data may obscure potentially interesting principles. In spite of the fact that such a general theory cannot be tested directly, each example shows that by making a detailed examination of variable articulatory data, we may be able to advance our thinking about theoretical issues.