Sensory Goals for Speech Movements: Cross- Subject Relations among Production, Perception and the Use of an Articulatory Saturation Effect

This paper describes two studies of 19 young-adult speakers of American English in which measures were made of: (a) phoneme contrast – articulatory distance between vowels in each of two vowel pairs and acoustic distance between the sibilants /s/ and //, (b) auditory discrimination of tokens from synthetic vowel and sibilant continua, and (c) contact of the tongue tip with the lower incisors for /s/. In the vowel study, an articulatory measure of contrast distance was correlated across subjects with auditory discrimination. In the sibilant study, acoustic contrast distance was related across subjects to auditory discrimination and also to the use of contact. These findings are compatible with the DIVA model of speech motor planning, in which the goals for phonemic speech movements are in auditory and somatosensory spaces. INTRODUCTION The two studies described below are based on the hypothesis that the goals of phonemic speech movements are in auditory and somatosensory spaces. Such phonemic goals are an essential characteristic of DIVA, Guenther’s model of speech motor planning (cf. Guenther & Ghosh, 2003; Guenther, et al., current proceedings), in which production and perception are closely linked. In order to probe this linkage, we have performed two studies of relations between production and perception. Recent brain imaging studies have provided evidence supporting the hypothesis of an intimate relation between speech production and speech perception. Investigators have shown that motor areas of the brain are active during speech perception (cf. Rizzolatti & Arbib, 1998) and auditory areas are active during speech production (cf. Hickok and Poeppel, 2000). The studies described below address this hypothesis in another way, by seeking correlations between measures of production and perception across speakers. Specifically, we hypothesize that speakers who discriminate well between phonetic stimuli with subtle acoustic differences will produce sound contrasts that are relatively clear-cut while speakers who discriminate less well between the same stimuli will produce less distinct contrasts. In DIVA, planning of movements to achieve phonemic goals is influenced by the listener’s need for clarity, competing with the speaker’s motivation to achieve an economy of effort (cf. Guenther, 1995; Lindblom, 1986). Clarity can be defined as the distinctness of auditory goal regions for different sounds, which is in turn determined by the size and location of those regions in auditory space. The model forms goal regions by monitoring the sounds of the language, and learning, for each phoneme, the region that encompasses all the examples of that sound. According to the model’s functionality, speakers who can perceive fine acoustic details will learn goal regions that are smaller than speakers with less acute perception, because speakers with acute perception are more likely to reject poorly produced tokens when learning the goal regions. STUDY 1: PRODUCTION AND PERCEPTION OF VOWEL CONTRASTS (Perkell et al., submitted-a) Production experiment The subjects for this and the subsequent studies were 19 young adult speakers of American English, 10 males and 9 females. Each subject participated in a production experiment in which we recorded Proceedings of the 6th International Seminar on Speech Production, Sydney, December 7 to 10, 2003.