Key points are not available for this paper at this time.
A maximum-likelihood procedure for estimating threshold in a yes–no task is presented. The absolute thresholds at six audiometric frequencies for a group of normal-hearing listeners were measured using this procedure. The average standard deviation of the threshold estimate is less than 5 dB with only 12 trials per threshold estimate. The total time taken to measure the audiogram was less than 3 min. The estimate of false-alarm rate is poor and strongly biased, but can be considerably improved by a simple modification that increases the testing time about 25%. Research supported by National Science Foundation and National Institutes of Health.
David M. Green (Thu,) studied this question.