Self-monitoring in speech
We all make mistakes in our speech. Fortunately, we are also equipped with a self-monitoring system, which inspects the quality of our speech and triggers interruption and self-corrections. According to Levelt's (1989) 'perceptual loop theory', monitoring one's own speech is like listening to other's speech -- the system that detects errors is the language comprehension system. But in the case of self-monitoring, there are two channels feeding into language comprehension: (1) one's own overt speech and (2) a representation of speech before it gets articulated (a phonological and/or phonetic code). In the first part of my talk, I will focus on the division of labour between these two monitoring channels. I will present a simple mathematical model which expresses the accuracy of the two channels as a function of empirically observable variables. I will use this model to test whether the two channels are equally accurate, as would be predicted by Levelt's theory. The second part of the talk focusses on the effects of 'covert' repair (repair of errors that are not yet articulated) on the fluency of speech. I will argue that Hartsuiker & Kolk's (2001) model of the time-course of self-monitoring can account for the distribution of disfluencies over different word types and positions in the utterance, as observed in a group of people who stutter.