Date of Award


Document Type

Open Access Thesis



First Advisor

Stanley Dubinsky

Second Advisor

Robin Morris


Disfluency is common in spontaneous speech. Self-correction is a type of disfluency that consists of reparandum, filler, and repair (Levelt, 1989). Little is known about the processing of self-corrections in a normally disfluent speech, and even less is known about its processing in atypically disfluent speech (e.g. speech in patients with autism spectrum disorder, hearing impaired, patients with brain damage, and stuttered speech; see: Lake, Humphreys, & Cardy, 2011; Lind, Hickson, & Erber, 2004; Plexico et al., 2010; Rossi et al., 2011; Yairi, Gintautas, & Avent, 1981). This study focuses on self-correction disfluencies in garden-path sentences and employs a behavioral data collection method to investigate how disfluencies are processed as they are heard. This experiment examines spoken language comprehension by measuring accuracy and response time to comprehension questions. The data was gathered and analyzed. Two experimental conditions were presented where in the first one normal speakers listened to typically disfluent speech, and in the second one normal speakers listened to atypically disfluent stuttered speech. The information about the speakers in the recorded stimuli was kept from the listeners. Fillers, such as uh and um are common in stuttered speech because of their helpful role in starting an utterance. In stuttered speech, the uhs, ums and pauses tend to be longer and in odd places, relative to the speech of people who do not stutter. Therefore, the hypothesis of this study was that the fillers and pauses made by people who stutter affect the dynamics of processing, particularly in garden-path sentences. Namely, the accuracy rate for the comprehensive questions was predicted to be lower for the garden-path filled pause sentences, particularly for atypical speaker condition. Reaction time was predicted to be longer for the same condition. The analysis revealed an accuracy measure dependence on the speaker condition but no significant time correlation. This study provides significant information about how normal speakers’ comprehension is affected by disfluency such as pauses in general, and how speech impairment, such as stuttering, affects the processing of filled and silent pause disfluecies.

Included in

Linguistics Commons