The relationship between inexperienced listeners' perceptions and acoustic correlates of children's /r/ productions

Abstract

This study examined inexperienced listeners' perceptions of children's naturally produced /r/ sounds with reference to levels of accuracy determined by consensus between two expert clinicians. Participants rated /r/ sounds as fully correct, distorted or incorrect/non-rhotic. Second and third formant heights were measured to explore the relationship between acoustic cues and perceptual judgments. Inexperienced listeners' agreement was greater for correct productions than for distorted or incorrect/non-rhotic productions. In addition, inexperienced listeners' differentiation of intermediate versus fully incorrect /r/ had lower sensitivity and specificity relative to an acoustically defined threshold than experienced listeners' classification. These findings are consistent with results of previous studies highlighting the difficulty in identifying gradations of correctness in misarticulated /r/, and they suggest that this ability may be influenced by clinical experience. Additionally, all listeners were noted to be more consistent in rating vocalic /r/ than consonantal /r/. Implications for clinician training and treatment planning are discussed.

Keywords:

• student clinicians
• vocalic /r/
• consonantal /r/
• distortions

Acknowledgements

Special appreciation is extended to the 12 New York University (NYU) students who were participants in the study. The authors also thank the Montclair students who assisted with acoustic measurements, Elizabeth Varall and Corey Silverstein, and Ryan Bennett at NYU who assisted with preparing the tokens to be used for the study. Finally, the authors thank Eugene Buder and Suzanne Boyce for their input on measuring formants in child speech. Aspects of this research were presented at the International Child Phonology Conference (2011) and the annual conference of the American Speech-Language Hearing Association (2011).

Declaration of Interest: The authors report no declarations of interest.

Notes

During the training period, students heard a small number of /r/ productions that were subsequently used as experimental stimuli in the /r/ rating task. Because the number of trained tokens was so small relative to the total sample size (12 out of 406) and the trained tokens were evenly distributed across rating categories (two each from the “1”, “2” and “3” categories), there is no reason to believe that prior exposure to these tokens influenced the rate of agreement across categories.