![Sample our Language & Literature journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5941/TOC-Subject-Sample-2021-Language-Literature.jpg"})
ABSTRACT
The aim was to determine how backward-looking sentence processing is affected by typically disfluent versus stuttered speech. Two listener groups heard Garden Path (GP) and control sentences. GP sentences contained no disfluency, a silent pause, or a filled pause before the disambiguating verb. For one group, the sentence preambles additionally contained stuttering-like disfluencies. Comprehension accuracy, event-related potentials (ERPs) time-locked to disambiguating verbs, and perceptual speaker ratings, were compared between groups. The With Stuttering group perceived the speaker as less competent but had better comprehension accuracy for GP sentences. ERPs to disambiguating verbs in GP sentences included a P600 component, indexing backward-looking sentence processing, but only for the No Stuttering group. Other ERP components, elicited to GP sentences with silent/filled pauses, did not differ between groups. Results suggest that listeners abandon prior expectations when processing sentences containing stuttering-like disfluencies, possibly because they lack a speaker model defined by the presence of stuttering.
Acknowledgements
We are grateful to Alissa Belmont for her efforts in preparing the stimuli, collecting and processing the data presented here. We are also grateful to the speaker who stutters who provided verbal stimuli for this study.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1. We acknowledge that speakers who stutter do not necessarily stutter on every utterance produced, although some might. For purposes of this study, stuttering was included in each sentence heard by the With Stuttering group to ensure that listeners in this group would unambiguously identify the speaker as someone who stutters.
2. It is important to note that sentence interpretations and, thus, responses to probe questions may be driven by pragmatic and syntactic processes operating in tandem (Bailey & Ferreira, Citation2003). Thus, in response to the GP sentence, “While the man hunted the deer ran into the woods”, listeners may answer Yes when asked “Did the man hunt the deer?” based on the pragmatic inference that the hunter was hunting the deer, even though syntactically the deer is not the object of the hunter. For all of the sentences used in this experiment, the first and second noun phrases were plausibly related, allowing for this same kind of pragmatic inference to be made. Thus, responses to probe questions were not treated as wholly correct or incorrect but, rather, as reflecting listener interpretations based on pragmatics and/or syntax. For this reason, all trials (whether interpreted correctly from a syntactic perspective or not) were included in the analysis of the ERP data.
3. For Control and GP sentences, the critical verb was preceded immediately by lexical material. For GP-disfluency sentences, the critical verb was preceded immediately by a filled or silent pause. Our concern was that the different material preceding the critical verbs in the different conditions elicited different types of ERP activity, unevenly affecting ERP amplitude immediately preceding and at the onset of presentation of our critical verbs in the different conditions. Other published studies have used a post-stimulus baseline correction procedure to control for this possibility (e.g. Neville, Nichol, Barss, Forster, & Garrett, Citation1991; Friederici, Hahne, & Mecklinger, Citation1996; Hahne, Citation2001; Hahne & Friederici, Citation1999; Hahne & Friederici, Citation2001; Osterhout et al., Citation1994; Maxfield et al., Citation2009). We adopted their approach. In order to validate this procedure, Friederici et al. (Citation1996) (among other examples) first baseline-corrected their averaged ERPs using a post-stimulus interval. They then determined, via statistical analysis, that amplitude differences were not present in the ERPs between conditions during the post-stimulus baseline interval. Along this same line, our Principal Component Analysis (reported below) failed to uncover any statistically significant ERP amplitude differences between conditions in the duration covering our post-stimulus baseline interval (0 to +100 ms), i.e. there were no temporal factors with a peak latency between 0 and +100 ms post-verb that captured ERP amplitude differences between conditions.