UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Evaluations of /r/ attempts of children in speech therapy by speech-language pathologists and child educators Radanov, Bosko


Background: Previous studies of treatment for English /r/ (designated with the North American symbol /r/) have mainly used Speech-Language Pathologists (SLPs) as expert listeners and scalar rating methods (e.g. Chaney, 1998). Tasks have involved rank order judgment of natural or synthesized speech stimuli, with a variety of trained and untrained adult and child listeners. Aims: The present study set out to compare expert and untrained listener evaluations of different /r/ attempts by children. The two comparison groups were SLPs and Educators (teachers or child care workers). A secondary objective was to compare an identification listening task with a paired comparison task. Methods and Procedures: Sixteen /r/ syllables ([ræ], [ar]) were extracted from pre- and post-treatment field recordings of four Canadian English-speaking children. The two tasks (identification of tokens as /r/ or not /r/, and a forced choice comparison of /r/ pairs) were presented through Microsoft Powerpoint under headphones. Twenty SLPs and eighteen Educators judged the quality of the /r/ attempts. Formant analyses were also made of the stimuli. Outcomes and Results: The expert listeners (SLPs) showed higher intra-rater reliability: 91% on the pairwise comparison task and 81% on the identification task, compared with 84%. and 78% for the untrained listeners respectively. Inter-rater reliability on single measures (ICC Educators=.51 in comparison, .21 in identification; SLPs=.42 in comparison; .31 in identification) was lower than that of average measures (ICC Educators=.96 in comparison, .87 in identification; SLPs=.95 in comparison; .92 in identification) Rank order of sample ratings as on- or off-target was similar between the two groups. The rankings matched the normative formant data for /r/ published in Guenther et al. (1999) and Flipsen et al. (2000, 2001) for the best tokens, with SLPs providing a ranking closer to the acoustic norms. Conclusions and Implications: Trained listeners appeared to be better able to identify nuances in /r/ quality, as confirmed by acoustic analysis of /r/ tokens. Intra-rater reliability was higher for SLPs despite greater disagreement among SLPs for single measures of inter-rater reliability. The paired comparison task had higher reliability scores than the identification task for both listener groups

Item Media

Item Citations and Data


For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use.