OBJECTIVE
To develop a set of Mandarin monosyllabic list for the goal to use as a standardized speech recognition assessment tool in China with sufficient validity, reliability and sensitivity.
METHODS
Thirty lists were designed based on the following criteria: efficiency, phonemic-balance, familiarity and coverage, while each list was designed corresponding to 25 monosyllables. These lists were read by a male broadcaster, recorded digitally and composed into compact disc. Our work consisted of three phases. Phase I: Sixty adults with normal hearing were recruited from Beijing to repeat as many syllables which they heard as possible. According to the randomized block design, 30 lists were presented with 6 intensities including -1 dB, 5 dB, 11 dB, 15 dB, 21 dB and 27 dB HL(speech). The lists and intensities were counterbalanced across all participants. Recognition scores in individual intensities for each list were calculated, and then logistic regression was utilized to fit Performance-Intensity (P-I) function. Two-way (list No. and Intensity) repeated measurement analysis of variance and Post-Hoc Tukey HSD test indicated that 22 lists were equivalent. Phase II: Twenty-two oral/aural normal adults were recruited to assess monosyllable recognition scores with the 22 equivalent lists at 10 dB HL(Speech), according to the Latin-Square design. Tests were administered twice for all participants with the same procedure and situation during 6 to 35 day intervals. The differences in scores (after a "rationalized" arcsine transformation) among 22 lists across over the two sessions is 9.3%, the data were collected from 22 participants, the measurement error was calculated by SD (standard deviation), the critical difference (CD) for test score improvement was 18.3% (determined as SD x 1.96, in 95% confidence level). Phase III: Eighteen participants with sensorineural hearing loss were recruited to assess recognition perception using 18 equivalent monosyllable lists at 30 dB suprathreshold based on Latin-Square design. Tests were administered twice by using the same procedure and situation within 1 to 16 day intervals. The same approach in Phase II was utilized to calculate SD (8.3%). The CD was calculated as 16.3% (in 95% confidence level).
RESULTS
A set of standardized Mandarin recognition assessment material had been developed and it consisted of 22 equivalent phonemic-balanced lists with 25 monosyllables each. Approximately, every single list took 2 minutes, and thus it might be appropriate for clinical assessment. The P-I functions reveal that the recognition threshold was (8.30 +/- 0.84) dB HL(speech) and the slope of PI functions was (4.0 +/- 0.3)%/dB for adults with normal hearing. When a set of Mandarin monosyllable lists was utilized as an assessment tool, the critical difference of 18.3% (for normal-hearing adults) and 16.3% (for hearing-impaired adults) would be a key for clinicians to assess the improvement of speech recognition ability appropriately with statistically significance.
CONCLUSION
In this study, a new Mandarin monosyllabic lists has been successfully developed with a sufficient validity, reliability and sensitivity for clinical evaluations, thus it might be convenience and helpful to be used as a standardized speech recognition assessment tool in China.