Robert F. Port
Department
of Linguistics,
Department of Computer Science
Program in
Cognitive Science
Indiana University, Bloomington, Indiana
47405
The phenomenon of incomplete neutralization and the subtlety of this incompleteness reveals vividly that speech sounds do not fall into discretely distinct phonetic types, and also that auditory impressions cannot be relied upon to identify them. AMR appreciates that these failures pose a serious problem for phonology. The reason is that the segmental units of standard phonology inherit the properties of discreteness and auditory identifiability from universal phonetics. So if phonetics is not always discrete, and practical identification is inherently unreliable, then phonology must restructure itself from the ground up. Phonology would then have no way to account for its discrete phonological objects (whether phonemes, rules, syllable trees, metrical grids, constraint rankings, allophonic rules, etc). This seems to be the theoretical bad dream that concerns AMR. I will argue that all this is true and that an explanation for the discreteness of phonology must be sought elsewhere, not in a hypothetical universal phonetic inventory. In fact, the explanation for phonological discreteness must be sought in the same place it is sought in other sciences (eg, astronomy, meteorology, economics, physics, biology, etc) - in the dynamically created structures often exhibited by systems with many degrees of freedom and a constant flow of energy.
Almost the same as Journal of Phonetics 24 (1996), 491-511, a Letter-to-the-Editor in response to A. Manaster-Ramer ``Letter from an incompletely neutral phonologist'', Journal of Phonetics 24.