The Development of Psychometrically Equivalent Cantonese Speech Audiometry Materials

Objective: The aim of this study was to develop and psychometrically evaluate speech audiometry materials that can be used to measure word recognition (WR) and speech recognition testing (SRT) in quiet for native speakers of Cantonese.

Study Sample: Commonly used bisyllabic and trisyllabic Cantonese words were digitally recorded by native male and female talkers and then evaluated by twenty normal-hearing Cantonese listeners.

Design: The recorded bisyllabic words were psychometrically evaluated and arranged into four WR lists and eight half-lists that are relatively homogeneous in audibility. Using logistic regression, SRT materials were developed by selecting 28 trisyllabic words with relatively steep psychometric functions and digitally adjusting their intensity to match the listeners' mean pure-tone average.

Result: The mean psychometric slopes for the WR materials were 7.5%/dB for the male talker and 7.6%/dB for the female talker, with no statistically significant differences between the lists or half-lists. At intensity levels required for 50% intelligibility, the mean psychometric slopes of the male and female talker SRT materials were 14.5%/dB and 14.9 %/dB, respectively.

Conclusion: High-quality digital recordings of Cantonese speech audiometric WR and SRT materials were developed and validated in this study. These materials are available on compact disc, indexed by talker gender.