How do listeners extract phonetic information from the speech signal? More than 50 years after the appearance of the motor theory of speech perception, recent neurophysiological discoveries challenge the view that speech perception relies on purely auditory mechanisms and suggest that the motor system might also be crucial for speech comprehension. The aim of the present chapter is to review these findings in an attempt to define what could be the “common language of perception and action”.

Understanding and modeling biological and physical mechanisms underlying speech production helps understanding speech motor control and its link to linguistic structure. Two of our recent studies are presented to illustrate this methodological idea. In the first study a realistic biomechanical tongue model is used to explore comprehensively the possible tongue shapes in the mid-sagittal plane. It is shown that the main directions of the tongue deformation observed in different languages are not related to any specific control, but rather to biomechanical and anatomical facts. The second study addresses the ‘trough effect’ observed in tongue shapes for bilabial consonants in VCV sequences. It is suggested that the combined analysis of kinematic and electromyographic data allows questioning previous interpretations of the trough effect with respect to speech motor control.

In order to accurately follow the growth of the vocal tract and the speech articulators and to infer their motor control and acoustic potentialities, it is essential to have anatomic and physiologic data throughout ontogenesis, from gestation to adulthood. The morphogenesis of the vocal tract, which involves the bony structures, their development, and cranio-facial growth during ontogenesis, is far from linear. The new data provided by genetics help interpreting the process of bony growth and thus of vocal tract reshaping from fetus to adult. To predict the consequences of vocal tract growth on the first speech productions (proto-vocalic and proto-consonantal articulations, babbling), anthropomorphic articulatory models are developed based on articulatory data. The articulatory models can generate realistic acoustic stimuli, which enable the testing of hypotheses about adult and newborn perceptuo-motor processes. These articulatory models can provide phylogenetical cues to the debate around the emergence of speech. The study of vocal tract growth therefore constitutes a key experimental paradigm for speech research.

Infants acquiring their native language have to learn, among other things, the words of this language. To this end, they have to extract word forms from the continuous speech stream, therefore they need to segment it. In a first section, we will show that several low-level cues, directly accessible through a surface analysis of the speech signal, may allow young infants to find words in sentences even before the end of their first year of life. In addition, infants have to assign a meaning to these phonological forms. In the second section, we will see that this task may be facilitated by syntactic knowledge. Once again, we will investigate the role of cues that are directly available in the speech signal, function words and prosodic cues.