Errors and strategy shifts in speech production indicate multiple levels of representation

Abstract

An earlier pilot study [Gick, J. Acoust. Soc. Am. 112, 2416 (2002)] observed that speakers producing challenging sequences of articulatory movements exhibit categorical shifts in production strategies. For example, a speaker may initiate all two-flap sequences with an upward tongue movement, then shift to a downward movement, then shift again. The present study attempts to determine how these strategy shifts can inform phonologists as to levels of representation. Ten subjects produced sequences of flap allophones of English /t/ and /d/ in an experimental paradigm similar to the pilot study. Results show strategy shifts in the speech of 7/10 subjects, such that: (1) shifts are more likely to occur following speech errors; (2) errors are more likely to occur in sequences with more flaps and more conflicts; (3) errors in response to articulatory conflicts occur at multiple levels of representation (segmental, lexical, etc.); (4) strategy shifts occur at multiple levels of representation (subsegmental, segmental, whole word); (5) word frequency does not affect strategy shifts. These results suggest that speech processing is active in parallel at multiple levels associated with phonological representation. [Work supported by NSERC.]