Allophony: Phonology or Phonetics? (talk recording available)
Dr. Mark Liberman
Tuesday, September 13, 2016, 01:00pm - 02:00pm
University of Pennsylvania, Departments of Linguistics and Computer and Information Science
The traditional organization of phonological theories involves a crucial redundancy, and serious consideration of this redundancy suggests a radical simplification of the theory. In technical terms, allophonic variation can be treated in at least two different ways: first, as a mapping from symbols to symbols, via phonological rules or constraints; or second, as a mapping from symbols to signals, via principles of phonetic realization. Careful examination of specific cases of allophonic variation generally suggests (and never seems to refute) a mode of description of the second type, in which structured phonological representations are mapped onto classes of phonetic trajectories. We should therefore consider the null hypothesis: a theory that entirely eliminates the symbolic treatment of allophonic variation, and makes post-lexical representations subject to direct phonetic interpretation, without any intervening symbol-manipulation, whether by rules or by constraints.
This leaves us with four well-motivated and indeed unavoidable tools for dealing with sound-structure patterns:
- Phonological inventory: The set of available phonological elements and structures.
- Lexical entries: The phonological spelling of whatever entities are listed in the lexicon: roots, affixes, morphological templates, words, phrases.
- Allomorphy: Alternative lexical pronunciations, whether conditioned by morphological features and morphological or phonological context, or in (linguistically) free variation.
- Phonetic interpretation: the mapping between symbols (from 1-3 above) and signals.
Given those resources, the phenomena generally described under the heading of allophonic variation do not require the addition of a fifth tool, in the form of manipulation of symbolic phonological representations via rules or constraints. Occam’s Razor therefore suggests a null hypothesis that some may find surprising, since it is inconsistent with many aspects of the past century of phonological practice: Phonological rules or constraints of the traditional symbol-manipulating sort do not exist.