Perceptual interactions and response bias in disordered voice assessment

University Of Cincinnati, Cincinnati OH

Investigators

Abstract

Project Summary Voice disorders are common and can have substantial negative impacts on quality of life. Diagnosis and treatment of voice disorders depend crucially on accurate assessment. Standard clinical tools include perceptual assessment scales (e.g., breathiness, roughness). These scales have questionable validity due to highly variable intra- and inter-rater reliability and agreement. Problems with scale validity hinder effective communication between clinical professionals and limit the quality of patient care. Evidence indicates that perceptual variability and response bias are important factors driving variability in the measured reliability and agreement of perceptual assessment scales. In addition, the relationships between voice acoustics and perceptual assessment scales are complex and multidimensional. The proposed research aims to adapt general recognition theory (GRT), a well-established multidimensional mathematical model of perception and response selection, for use as a measurement instrument for disordered voice assessment. A series of perceptual experiments will be conducted, and the data will be analyzed with multilevel GRT in order to measure the strength of the relationships between two acoustic properties of voice and two clinical assessment scales, while simultaneously measuring perceptual variability, perceptual interactions, and response bias. The two acoustic dimensions of interest are H1-H2, the difference in amplitude dB of the first and second harmonics of a voice, a measure of spectral shape at low frequencies, and HNR, the harmonic-to-noise ratio, a measure of the relative strength of the harmonic and inharmonic components of voice. The two assessment scales of interest are breathiness and roughness. Data will be collected for each of two assessment scale designs, one with four-point ordinal responses, the other with continuous visual analog and direct magnitude estimation responses. In addition to describing the perception and response selection underlying each individual listener's responses, the proposed model will enable quantification of between-listener variability with respect to the perception of disordered voice acoustics, perceptual variability, perceptual interactions, and response bias. The proposed research will establish a powerful quantitative framework for investigating voice perception and clinical voice assessment, and the proposed measurement model will enable more effective communication between clinical professionals, improve patient care, and strengthen the statistical foundation of voice perception research. The proposed work will also lay the foundation for future research on clinical voice assessment and the perception of voice in the telehealth delivery model, in which external noise and signal degradation may introduce additional complications.