The aim of this research is to describe the development of an adaptive Australian Sentence Test in Noise and to validate the test in terms of test–retest reliability and efficiency using data obtained from its clinical application.
The relative intelligibility of 1264 Bamford-Kowal-Bench (BKB)–like sentences in the presence of competing four-talker babble was assessed with cochlear implant recipients. Intensity adjustments to the babble segments were made to reduce intersentence variability. Computer software was developed to administer an adaptive speech reception threshold (SRT) test using these adjusted sentence/babble pairs and test–retest SRT data from a separate group of 23 cochlear implant recipients was analyzed, comparing different SRT calculation and test stopping rules.
The adjusted sentence/babble pairs were used in clinical studies to obtain an SRT by presenting 32 sentences. Analysis of test–retest pairs of SRT data from 23 recipients indicated that a psychometric fit SRT calculation rule provided better reliability than did the Hearing in Noise Test (HINT) calculation rule, or rules based on mean turns. This rule, using the morpheme correct scores for each sentence, gave a standard deviation for a single SRT of 0.76 dB. Further analyses revealed that the test could be shortened to 20 sentences with an increase of 0.19 dB in variability, while reducing the median test time by approximately 2 min.
This article reports validation data for a new Australian Sentence Test In Noise. When 20 BKB–like sentences are used with a psychometric fit calculation rule, a standard deviation of approximately 1 dB is obtained in approximately 3 min 36 sec.