Trends Hear. 2022 Jan-Dec;26:23312165221134007. doi: 10.1177/23312165221134007.
A new sentence recognition test in Mandarin Chinese was developed and validated following the principles and procedures of development of the English AzBio sentence materials. The study was conducted in two stages. In the first stage, 1,020 sentences spoken by 4 talkers (2 males and 2 females) were processed through a 5-channel noise vocoder and presented to 17 normal-hearing Mandarin-speaking adults for recognition. A total of 600 sentences (150 from each talker) in the range of approximately 62 to 92% correct (mean = 78.0% correct) were subsequently selected to compile 30, 20-sentence lists. In the second stage, 30 adult CI users were recruited to verify the list equivalency. A repeated-measures analysis of variance followed by the post hoc Tukey’s test revealed that 26 of the 30 lists were equivalent. Finally, a binomial distribution model was adopted to account for the inherent variability in the lists. It was found that the inter-list variability could be best accounted for with a 65-item binomial distribution model. The lower and upper limits of the 95% critical differences for one- and two-list recognition scores were then generated to provide guidance for detection of a significant difference in recognition scores in clinical settings. The final set of 26 equivalent lists contains sentence materials more difficult than those found in other speech audiometry materials in Mandarin Chinese. This test should help minimize the ceiling effects when testing sentence recognition in Mandarin-speaking CI users.