Developing and testing descriptive models of English pronunciation

Background:  In my opinion, modern American theoretical linguistics has been dominated by untested hypotheses which have significantly weakened the ability of researchers to provide effective and practical means for helping people learn languages.  The notions that pronunciation will develop automatically as people work on communicative activities, or that grammar does not have to be taught because people will pick it up naturally, have undermined important areas to which CALL can make significant contributions.

Research question:  What are acceptable ranges for acoustic parameters of different registers of American English pronunciation, including consonants, vowels, and suprasegmental features?  Along with this come questions such as “To what extent is English really a stress timed language?”  This provides for a much more dynamic approach to phonological structures.

Suggested methodology/comments:  Model creation could take the form of gathering representative samples of spontaneous and scripted speech from a variety of first language backgrounds and English, analyzing the samples acoustically, entering the parameter ranges in spreadsheets, comparing them and creating the model.  Model testing would consist of analyses of data from new subjects who were not part of the model building process.

Contact:  Garry Molholt   Gmolholt@WCUPA.EDU

Reader Comments: --