Phonotactic experience conditions speech perception

Abstract

Numerous studies have shown that native phonotactic constraints influence listeners’ perception of non-native speech. These studies typically show that speech which violates the phonotactic rules of the listener’s native language is often misperceived as speech that adheres to native phonotactics, behaviour which is often characterised as phonotactically conditioned perceptual repair. The present thesis extends these findings by examining the effects of language-specific listener expectation on non-native speech perception. The central hypothesis presented within is that sequences of speech that are unexpected by the listener are often misperceived as sequences that are expected. This suggests that phonotactically conditioned perceptual repair is not necessarily caused by an innate understanding of the listener’s native phonotactic constraints: sequences of speech that contain phonotactic violations may be considered as one extreme of a continuum of expectedness. Additionally, the present thesis explores the relationship between predictability and attention. The studies presented show that listeners pay less attention to incoming speech input when they can make predictions regarding upcoming input based upon preceding materials. The present thesis tests these hypotheses in six studies conducted on native Japanese speakers. These experiments use allophonic variability, transitional probability, and word frequency as measures of expectation. The results confirm the central hypothesis, they show that listeners misperceive sequences of speech that are improbable in Japanese. The way that listeners misperceive speech is also found to be dependent on expectation, when presented with unexpected speech that may be repaired in multiple ways, listeners experience the more probable illusion.