The development of sentences for speech-in-noise test in Tamil for children, younger adults, and older adults

Abstract

Introduction: The present study aimed to develop and standardize sentences for speech-in-noise tests in the Tamil language for children, younger adults, and older adults. Material and Methods: A set of 300 sentences in Tamil was taken from the “Production of Language Training Materials in Major Indian Languages” (UNICEF-funded; Manjula, Karanth, Geetha, & Prema, 2010) and recorded. These sentences were rated by five native Tamil-speaking speech-language pathologists based on naturalness, predictability, and identifiability. Sentences highly rated for naturalness, low predictability, and high identifiability were selected by at least 80% of the raters, resulting in 147 sentences chosen for further use. These sentences were mixed with speech-shaped noise at varying signal-to-noise ratio (SNR) levels, ranging from +5 to -10 dB in 2.5 dB steps. The speech perception in noise ability was assessed in 90 normal hearing participants (30 children, 30 younger adults, and 30 older adults). Results: Perceptual SNR-50 values were calculated for each list, based on the perceptual score obtained by each participant, separately for children, younger adults, and older adults. And statistical analysis determined that some lists had significantly different perceptual scores and were excluded. Seven lists were finalized for each age group, with mean SNR-50 values of -4.7645 dB for children, -4.6628 dB for younger adults, and 4.6549 dB for older adults. Reliability and internal validity tests confirmed the reliability and validity of the test for assessing speech perception in noise abilities across different age groups.Conclusion: The study found that speech identification scores decreased with lower SNR levels, with significant differences in SNR-50 values across children, younger adults, and older adults. Separate, equivalent test lists were created for each age group. The test materials showed high test-retest reliability and internal validity, allowing them to distinguish between individuals with normal hearing and those with hearing loss based on SNR loss.