Résumé
Since Lieberman and Crelin (1971), the question of vocal tract abilities and the link between anatomy and control has been the object of a number of conflicting papers. Part of the debate concerns the acoustic possibilities of the Variable Linear Articulatory Model (VLAM), an articulatory model that has provided the foundation of our own work for many years. VLAM is considered by Lieberman and some others as misleading because of its supposed overestimation of phonetic capabilities of human newborns. In this paper, we compare the VLAM simulations between 0 and 5 years with acoustic data on infant and child vocalizations from a number of studies in the literature. We show that the agreement is globally quite good, with no hint of overestimation above the age of 6 months for first formant and 15 months for second formant, while on the contrary simulations assessing the hypothetical role of proportions in an angled vocal tract with another model clearly diverge from ground truth child data. We conclude that limitations in infancy are a matter of control rather than anatomy. Then we lay a framework to situate "efficient acoustic modulation" within speech communication in general. We propose that the Frame-Content (FC) Theory by MacNeilage and Davis (2000) provides the basis of a vertical first component of a "principle of efficient modulation," giving birth to manner of articulation. We further propose that constriction control is the basis of the horizontal second component of efficient modulation, giving birth to place of articulation. These linked components provide a valid foundation for exploring the development of human vocal tract anatomy and control, now in two dimensions. We close by summarizing our own perspective on the possible role of swallowing in the evolution of this control, as a possible extension of the role of mastication in FC. (C) 2013 Elsevier Ltd. All rights reserved.