A simulation analysis of the variability of the roving level hearing test.

In the study of auditory prostheses, the Speech Recognition Threshold (SRT) is the Signal to Noise Ratio (SNR) at which 50% of words are correctly identified. SRT is typically measured using an adaptive procedure wherein speech is presented at a fixed sound pressure level (SPL) and the noise level is varied according to the subject's responses. A roving level SRT test has been used by researchers with the goal of including the effectiveness of Automatic Gain Control (AGC) systems in SRT measurements. The roving method presents speech at three different SPLs with the level for each sentence chosen pseudo-randomly, while adaptively varying the SNR. This study used simulations to compare roving and fixed level SRT tests. It was found that roving level tests have significantly increased variability when there are level-dependent differences in subject scores. The interleaved level test is recommended as an alternative as it provides clear visibility of level-dependent performance and a better understanding of overall subject performance.