Andries Coetzee recently delivered a keynote address at the International Conference on Phonetics and Phonology, held from September 25-27 on the campus of Keio University, Tokyo. This conference brought together scholars from across Japan and the rest of the world who work phonetics and phonology, with a specific focus on phonetic and phonological variation, and on word and sentence prosody. Other keynote speakers at the conference included Bruce Hayes (UCLA), Caroline Féry (Goethe University, Frankfurt), and Sun-Ah Jun (UCLA). Andries presented a talk on phonological variation, focusing on the computational modeling of the interaction between grammatical and non-grammatical factors. The title and abstract of his presentation are given below.
Modeling Phonological Variation
Existing approaches to phonological variation differ in the role that they ascribe to grammar. Some assume no role for grammar, considering variation as the result of non-grammatical factors impacting the categorical output of phonological grammar. Others assume that grammar and non-grammatical factors contribute equally to variation. Yet others attempt to account for all aspects of variation with grammar alone, allowing no room for non-grammatical factors. In this talk, I will develop a fourth possible model with the following features: (i) Grammar itself is variable and hence contributes to variation. (ii) Non-grammatical factors also contribute. (iii) But the model is grammar dominant – grammar defines the space of possible variation and non-grammatical factors can only influence how variation is realized within this grammar-delimited variable space. The model will be implemented in Harmonic Grammar, and will be applied to English variable t/d-deletion (west bank~wes_ bank),variable cross-word nasal place assimilation (green boat~greem boat), and variable devoicing of Afrikaans plosives, focusing on the interaction of grammar with factors such as usage frequency, speech rate, and communicative context.