The dispersion-focalization theory of sound systems

Abstract

The Dispersion-Focalization Theory states that sound systems in human languages are shaped by two major perceptual constraints: dispersion driving auditory contrast towards maximal or sufficient values [B. Lindblom, J. Phonetics 18, 135–152 (1990)] and focalization driving auditory spectra towards patterns with close neighboring formants. Dispersion is computed from the sum of the inverse squared inter-spectra distances in the (F1, F2, F3, F4) space, using a non-linear process based on the 3.5 Bark critical distance to estimate F2’. Focalization is based on the idea that close neighboring formants produce vowel spectra with marked peaks, easier to process and memorize in the auditory system. Evidence for increased stability of focal vowels in short-term memory was provided in a discrimination experiment on adult French subjects [J. L. Schwartz and P. Escudier, Speech Comm. 8, 235–259 (1989)]. A reanalysis of infant discrimination data shows that focalization could well be the responsible for recurrent discrimination asymmetries [J. L. Schwartz et al., Speech Comm. (in press)]. Recent data about children vowel production indicate that focalization seems to be part of the perceptual templates driving speech development. The Dispersion-Focalization Theory produces valid predictions for both vowel and consonant systems, in relation with available databases of human languages inventories.