Development of an Afrikaans test for sentence recognition thresholds in noise

Abstract

OBJECTIVE: The development of a valid and reliable Afrikaans test of sentence recognition thresholds in noise. DESIGN: A collection of sentences was developed, rated for naturalness and grammatical complexity, and digitally recorded using a female speaker. Sentences found to have similar psychometric curve slopes, with equivalent intelligibility at three different noise levels, were arranged into 22 phonemically matched lists of ten sentences each. List equivalence was evaluated in normal-hearing listeners in full and reduced bandwidth conditions. Test-retest reliability of the remaining lists was evaluated in a second group of listeners. STUDY SAMPLE: All listeners were native speakers of Afrikaans with normal hearing. For evaluation of list equivalence, ten listeners were used. Twenty other listeners were used to evaluate test-retest reliability. RESULTS: A collection of eighteen phonemically matched lists was produced. Lists were found to be of equivalent difficulty in full and reduced bandwidth conditions, and to have good test-retest reliability in normal-hearing listeners. The average recognition threshold of these lists was -2.73 dB signal-to-noise ratio (standard deviation = 0.64 dB), and within-subject variability was 1.22 dB. CONCLUSIONS: The developed test provides a valid and reliable means of measuring sentence recognition thresholds in noise in Afrikaans.