Evaluation of Stopping Rules for Audiological Ascending Test Procedures Using Computer Simulations

Stopping rules for ascending audiological test procedures were evaluated by Monte Carlo simulation. The stopping rules differed in the minimum number of responses required at a level, whether these responses occurred on half or a majority of the ascending series, and whether all or only the most recent ascending series were considered. For all procedures, shallow psychometric function slopes yielded estimated thresholds closer to the level giving 50% detection than did steeper slopes, which yielded estimated thresholds roughly 2.5 dB above the level corresponding to 50% on the psychometric function. Shallow slopes also resulted in decreased reliability across threshold measurements, an increased number of trials required for threshold estimates, and a higher proportion of measurements that had to be repeated. Stopping rules using a two-response criterion were faster than those using a three-response criterion, with only a small decrease in reliability. Among stopping rules using the same number of responses for criterion, differences were seen primarily in efficiency for shallow slopes, particularly for procedures using a three-response rather than a two-response criterion for stopping. Results from these simulations should be useful to standards groups.