Predicting the Structure of Phonological Systems

Abstract

Abstract Do the phonological predictions derived from phonetic models impose restrictions on the definition of’possible phonological system’? This question is addressed by examining the spectral integration model of vowel perception. The output of the model is investigated under a series of phonetic manipulations and translated into a set of phonological predictions, which are then compared against cross-linguistic patterns in vowel processes and inventories. It is argued that, despite the serious constraints on predictive power imposed by language-specific aspects of phonetic and phonological structure, phonetic models such as the spectral integration model enable us to define the set of phonetically motivated phonological phenomena.