Spanish Vowel Sandhi: Toward a Characterization of Low-Level Processes in Phonology

250 p.Thesis (Ph.D.)--University of Illinois at Urbana-Champaign, 1984.In the gap between phonology and phonetics lie the low-level processes: highly variable sound changes of natural language which are crucial to decoding by the hearer, but in the speaker are more a product of performance than competence. Previous phonological frameworks have ignored such phonetic detail as "uninteresting" or have compartmentalized it into optional rules, variable rules, rules with numerical coefficients on features, speech tempo specifications on rules, etc. Much of this work ignores the full range of parameters in low-level variation or notationally conflates their complex relationships into pseudo-generalizations.Using Spanish vowel sandhi as an example, I survey six previous studies and discuss their attempts to relate the infinite variation of the data to their theoretical frameworks. Seeking a better model, I examine Mexican Spanish vowel sandhi in a corpus of spontaneous speech and conclude that the variation can be organized into types of continua based on "competing tendencies" like hiatus, gliding, shortening, centralization, elision, and coalescence.Three major classes of parameters are investigated for their roles in predicting specific types of phonetic realizations: (1) site parameters, (2) segmental parameters, and (3) production parameters. In my investigation of the effects of these parameters, I conclude that low-level variation results from the distribution of physiological energy expended in the function of the speech apparatus over the phonological segment string defined by language-specific message coding.For Spanish vowel sandhi, non-segmental boundaries affect the range of variation. The serial ordering of segments and their articulatory and acoustic targets determine potential tendency continua configurations. Production parameters such as stress, intonation, rhythm, and topic emphasis then interact to determine phonetic realizations on these continua.Using spectrographic analysis, I conclude that speech tempo is an overriding parameter in low-level variation, affecting in the data I examine not only the extent to which vowel sequences collapse but also the extent to which they are susceptible to other low-level effects of their segmental environments.A model for the characterization of Spanish vowel sandhi is suggested which may be applied to other low-level processes as well.U of I OnlyRestricted to the U of I community idenfinitely during batch ingest of legacy ETD