Date of Defense


Date of Graduation



Speech Pathology and Audiology

First Advisor

Stephen Tasko

Second Advisor

Ron Kelley


Purpose: The speech of persons who stutter can often sound unnatural. This can be due to a variety of factors including the presence of stuttering itself and/or the unintended consequence of speech-altering treatment. Understanding factors that contribute to speech naturalness within the stuttering population can direct improved management approaches. This study examines associations between speech task, behavioral measures of stuttering, and selected acoustic measures on listener ratings of speech naturalness for a group of adults immediately before and after participation in an intensive fluency-shaping treatment program. A secondary aim of this study is to assess the use of a novel rating scale (i.e. a visual analog scale) to judge speech naturalness.

Methods: Participants include 50 young adult college students who passed a standard hearing screening. The participants served as judges who rated the level of perceived speech naturalness of video recordings made on adults who stutter. A computer interface allowed judges to rate speech naturalness by moving a slider on a computer screen along a continuum ranging from highly natural to highly unnatural. The video recordings that were judged consisted of one-minute excerpts drawn from 40 adults who stutter performing an oral reading and monologue before and following participation in a group-based intensive, four-week fluency shaping program. Mean naturalness ratings were compared across the speech task (oral reading vs. monologue) and time (pre-treatment vs. post-treatment). Additionally, regression analysis was used to evaluate the relationship between speech naturalness ratings and measures of stuttering frequency and selected acoustic measures.

Results: Rater agreement using the visual analog approach to speech naturalness rating was considered excellent and was highly correlated with data previously collected using a more conventional nine-point equal appearing interval scale. A mixed-effects model was used to assess effects of speech task and treatment participation on mean speech naturalness ratings of the video samples. There was a significant main effect for treatment but not for speech task. Post-treatment samples were rated as more natural than the pre-treatment samples. Pre-treatment naturalness ratings were positively correlated with stuttering frequency, but post-treatment naturalness ratings were unrelated to stuttering frequency. Post-treatment naturalness ratings were negatively correlated with articulation rate, and positively correlated with mean pause duration.

Conclusion: Excellent rater agreement and strong correlations with more traditional rating scales suggest that a visual analog scale can be a viable scaling method for rating speech naturalness. Although stuttering frequency was a strong predictor of speech naturalness prior to treatment, it provided little prediction of speech naturalness following treatment. Instead, articulation rate and pause duration measured from the speech samples predict post-treatment naturalness ratings. Specifically, those speech samples with the slowest articulation rates and the largest speech pauses between phrases were associated with less natural sounding speech and may represent negative consequences of stuttering treatment. This study demonstrates that the combined use of stuttering frequency measures and acoustic measures of speech performance can help understand the factors that determine why some persons who stutter sound more or less natural than others.

Access Setting

Honors Thesis-Open Access