Psychoacoustic properties of imagined speech stimuli

Abstract

Synthetic consonant-vowel syllables were generated by factorially crossing four values each on two acoustic continua (VOT and F2-F3 transition onset frequency) to produce 16 stimuli identifiable as /ba/, /pa/, /ta/, and /da/. In addition, four silent stimuli were generated. The complete set of 20 stimuli was presented to subjects in a paired comparison test for scalar judgments of similarity. Silent stimuli were presented paired both with sounded stimuli and with other silent stimuli. On the response sheet sounded stimuli were indicated by an underscore and silent stimuli by “ba,” “pc,” “da,” or “ta,” depending upon which syllable the subject was supposed to “hear.” Subjects were instructed to make comparisons involving silent stimuli by imagining themselves pronouncing the silent syllable. Responses from 13 subjects were analyzed using the INDSCAL individual differences scaling approach [J. Carroll, in Multidimensional Scaling: Theory and Applications in the Behavioral Sciences, edited by Shepard, Romney, and Nerlove (1972)]. The recovered two-dimensional perceptual space clearly showed the four silent stimuli located appropriately with respect to the sounded stimuli, indicating that subjects were capable of making linguistic judgments of sounded and silent stimuli alike. These results provide direct support for prototype models of speech perception.