Upward bias in audiometric thresholds is caused by stopping at the first opportunity

Abstract

Tone-detection thresholds are the primary index of hearing. Although standard audiometry nominally estimates the stimulus level corresponding to 50% detection, thresholds are biased upward to an average of 98.1% at 500 Hz [Marshall, L. & Jesteadt, W. J Speech Hear Res 29, 82–91. (1986)]. This has been explained by the large difference in the probability of detection between stimulus levels near threshold. However, bias is greater than predicted for measured psychometric functions and does not change proportionally with step size. In standard audiometry, the stopping rule is defined with flexibility for clinical judgement. Our hypothesis is that upward bias is caused by stopping at the first opportunity, because it makes specific ordered combinations of hits and misses impossible, like hit-hit-miss. Monte Carlo simulations were used to compare audiometric thresholds obtained using different stopping rules. Thresholds obtained when stopping at the first opportunity had greater upward bias compared to a fixed-trial stopping rule independent of the total number of trials presented, as predicted. When a baseline test is biased upward by stopping at the first opportunity, a follow-up test may inadvertently miss a significant change in hearing by continuing to present stimuli after the first opportunity to stop. [Work supported by R01DC015051.]