Simulated audiometry reveals multiple asymptotic thresholds for the same listener depending on the number of presentations at threshold

Abstract

Changes in audiometric thresholds are used to detect auditory insult resulting from exposure to hazardous noise, blasts, and ototoxicity. Stimuli presented during audiometry reflect a stochastic process driven by listener responses, and thus a threshold can be reported after a different number of presentations, e.g., two hits and zero, one, or two misses. We interpret thresholds as equivalent despite large differences in the proportion of hits observed. To test for systematic bias in thresholds obtained after a different number of presentations, a continuous distribution of audiometric threshold as a function of stimulus level was numerically approximated for a fixed listener model tested using 2 and 5 dB ascending steps. Results were sorted by the number of presentations at threshold and summary statistics were calculated directly from the proportion of Monte Carlo simulations reporting each stimulus level as threshold. Thresholds differed systematically as a function of the number of times the threshold stimulus level was presented, consistent with a model combining the probability of meeting the stopping criteria with the asymptotic distribution of ascending trials. The implication is that recording the number of presentations at threshold on the audiogram will improve our ability to detect changes in hearing. [R01DC015051.]