汉语普通话单音节测听表的建立与评估

Development and evaluation of standardized Mandarin monosyllable audiometric materials

目的建立一套普通话单音节测听表并验证其效度、信度和敏感度，期望成为国内标准化的言语识别率测听材料。方法遵循简短、音位平衡、常用、覆盖等原则，生成30张音位平衡单音节表，每表对应25个常用汉字。由专业男性播音员朗读，经数字化录音及声学处理，录制成CD。研究分为三个阶段：①招募60名听力正常年轻人，采用随机区组设计：言语测听表号与测听强度均衡排布，每一受试者均变换6种强度（言语听力级-1、5、11、15、21、27dB）以循环顺序聆听全部30张单音节表。计算每表每一声级下的言语识别率得分，以logistic曲线回归来拟合识别率-强度（P—I）函数。对30张词表的识别率进行双因素（表号、强度）方差分析和Post—Hoc两两比较，得到22张等价表。②北京22名听力正常年轻人，采用拉丁方设计，在言语听力级10dB强度下依次测试22张等价表的识别率。间隔6～35d，在完全相同的条件下重新进行第二轮识别率测试。两轮测试的识别率经“合理化”的反正弦变换后，逐表汇总两轮测试分差，该分差值在22名受试者中的标准差为9．3％，转化成95％置信度下的临界分差为18．3％。③采用拉丁方设计方案，对18名感音性听力损失患者在其纯音平均听阈上30dB或最大舒适级，测试其对18张等价表的识别率。间隔1～16d在相同条件下再次进行测试。两轮测试的识别率经“合理化”的反正弦变换后，分别为（68．0±15．9）％和（69．2±14．6）％，配对t检验，P=0．107。两轮测试分差，在18名患者中的标准差为8．3％，转化成95％置信度下的临界分差为16．3％。结果创建了一套22张等价的标准化的言语识别率测听表。每表由音位平衡的25个单音节构成，播放时间约2min。听力正常人的单音节识别阈为言语听力级（8．30±0．84）dB，P—I函数的斜率为（4．0±0．3）％／dB。临床评判受试者言语识别率差异时，只要分差超过临界分差——正常人（18．3％）、患者（16．3％），即可确认具有统计学意义上的显著性差异。结论成功建立起一套普通话单音节测听表，具有较高的效度、信度和敏感度，便于临床应用。

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 HLspeech. 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 HLspeech, 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 phonemicbalanced 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 HLspeech and the slope of PI funtions 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.