This study argues that linguistically-determined aspects of phonetics form part of grammatical theory in much the same sense as phonology or syntax do, and can be modeled in terms of similar principles. It proposes that the phonetic component of grammar contains sets of phonetic representations similar to the partially specified, multitiered representations of the phonological component. At the acoustic level, these representations include a set of autosegmental tiers specifying values for acoustic parameters such as voicing, nasality, vowel formants, FO, etc. as well as a duration tier which organizes these values into a succession of discrete acoustic events in the time domain. This integrated representational system, or IRS, defines a complete interpretation of surface phonological representations at the acoustic level, and can be specified with sufficient detail to provide input to an acoustic speech synthesizer. This framework is applied to a study of the temporal properties of long vowels and diphthongs in General American English. We consider the well-known question whether the long vocalic nuclei of words like bait, boat, bite, and bout should be analyzed as one phonological segment or two. Durational evidence involving asymmetries in the distribution of contextually-determined lengthening across the syllable suggests that the nucleus of bait is best modeled as a single melodic unit (root node) occupying two skeletal positions, while that of bite is best modeled as two melodic units. The phonetically diphthongized quality of the nuclei of words like bait and boat is determined at the phonetic level by a constraint requiring them to have separate formant target positions at their left and right edges; as a result, no general long vowel diphthongization rule is required in the phonology.

This paper is copyrighted, and is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0) - see https://creativecommons.org/licenses/by-nc-nd/4.0/