Roles of speech errors, monitoring, and anticipation in the production of normal and stuttered disfluencies

In their Covert Repair Hypothesis (CRH), Postma and Kolk (1993) proposed that stuttering-like disfluencies arise, in both normal and stuttered speech, as a consequence of speakers‟ attempts to repair phonological-encoding errors before they start to speak. They posited that stutterers are particularly disfluent because they make larger numbers of such errors compared to normally-fluent speakers. To date, however, experimental research has provided little reliable evidence to support or counter this hypothesis. This thesis constitutes a systematic attempt to provide such evidence. Using a tongue-twister paradigm in conjunction with manipulations of auditory masking, it first documents (a) the vigilance with which normally-fluent speakers monitor for such errors; (b) the relative accuracy with which they detect them; and (c) the frequency with which they occur – in both inner and overt speech. A second set of experiments then extends the same investigation to a group of stutterers and matched controls and explores the relationship between the occurrence of participants‟ errors in the experimental paradigm and the frequency of their stuttering-like disfluencies in everyday speaking situations. Together, these experiments reveal that, compared to controls, participants who stutter monitor their speech with similar levels of vigilance; identify phonemic errors with similar degrees of accuracy; and, as predicted by the CRH, produce significantly more errors – in both their inner and overt speech. However, contrary to the predictions of the CRH, no relationship was found between the frequency of such errors in inner speech and the severity of participants‟ disfluencies. In a final set of experiments, a speech-recognition paradigm is employed to explore an alternative hypothesis: that stuttering-like disfluencies can be precipitated, in a speaker, by the mere anticipation that his words will result in communication failure. Results revealed that, for stutterers, stuttering decreased on words that were consistently followed by feedback implying correct recognition, but not on words followed by feedback implying incorrect recognition. For normally-fluent speakers, equivalent correlations were not found. The thesis concludes that slow or impaired phonological encoding may play a role in the development of the disorder. But, once established, the anticipation of communication failure may be a more important factor in determining where and when stuttering-like disfluencies actually occur. It then discusses implications of the experimental findings for hypotheses that posit a connection between phonological encoding and stuttering.