Development of Bisyllabic Speech Audiometry Word Lists for Adult Malay Speakers
Standardised speech audiometry material is essential in assessing hearing for speech; however, material in Malay language, particularly for speech reception threshold test, is limited and not thoroughly validated. This thesis examines the development of standardised, phonemically-balanced bisyllabic Malay speech reception threshold (SRT) test word lists for Malay-speaking adults. The effect of having a mixture of familiar and nonsense words on speech recognition is also explored. The processes of developing the word lists include selecting and compiling the words using content analysis research method, testing for homogeneity and consistency and validating the acoustic content, both using correlational research method, and assessing the clinical validity using concurrent validity method. The familiar words were selected from a corpus of familiar words extracted from daily newspapers while the nonsense words were formed based on linguistic properties of Malay. The preliminary set consisted of fifteen lists with 10 familiar words and 5 nonsense words in each. The analyses of the findings show consistency of speech discrimination using the word lists using Friedman test to have statistically no significant difference in correct scores achieved using any of the word lists, Χ2 = 19.584, p>0.05. Homogeneity test for all lists using Cronbach’s alpha showed a value of 0.78, indicating a strong agreement and good homogeneity among the lists. When five lists with inter-item correlation ≤0.8 were excluded from the homogeneity analysis, the alpha value for the remaining 10 lists increased to 0.88. Consistency analysis of acoustic content using repeated measures ANOVA showed no significant difference between the list and the LTASS, F=1.229, p>0.05. All 15 lists were then tested for clinical validity. Two versions of list content were assessed, an all-words version (AWL) containing all 15 words each list, and a meaningful-words only version (MWL) containing 10 meaningful words for each list. Correlation analyses between half peak level (HPL) of the speech recognition curve and pure tone (PT) thresholds showed that, in consideration of both normal hearing and hearing impaired listeners, the HPL correlated best with PT average of 250, 500, 1000, 2000 and 4000 Hz for both AWL (r = 0.67 to 0.95) and MWL (r = 0.65 to 0.95). A comparison between HPL and PT average of 250, 500, 1000, 2000 and 4000 Hz showed mean differences of 4 dB (SD = 3) and 3 dB (SD = 4) with the range of tolerance (95% confidence) of ±7 dB and ±8 dB for AWL and MWL respectively. Sensitivity, specificity, and positive and negative predictive values, when set at tolerance level of ±10 dB, were mostly >0.90 for normal hearing and hearing loss listeners using either versions. It was concluded that the addition of nonsense words does not significantly affect SRT. The correlation between the SRT obtained using the bisyllabic Malay word lists and the PT thresholds suggested that the word lists were robust enough to be used in assessing speech hearing clinically. In conclusion, the current study has achieved to develop and produce a standardised, phonemically balanced bisyllabic Malay speech audiometry (BMSA) word lists for assessing speech reception threshold and discrimination in adult Malay speakers.
- PhD