No abstract

No abstract

No abstract

At the last meeting of the Acoustical Society of America, a new method of estimating slopes of psychometric functions when thresholds were varying over the course of an experiment was reported. Computer simulations were used to calculate the relationship between the true slope of the function and the variability across pairs of trials in two interleaved adaptive tracks. Further, the variability across all trials in the two tracks was shown to indicate the stability of the function over time. Here, the validation of that procedure in the measurement of human performance is reported. Listeners were asked to detect a 1000-Hz tone in a narrow band of noise, the level of which was either fixed or varied sinusoidally from trial to trial. This masker variation imposed systematic threshold shifts analogous to those in the simulations. The period of sinusoidal shift ranged from 16–128 trials per cycle, and the amplitude of shift was either 4 or 8 dB. Psychometric functions for individual listeners were constructed both by traditional methods of fitting the trial-by-trial data, and by using the functional relationship between slope and variability determined from the computer simulations. When the imposed threshold shift did not occur too rapidly, the newer method provided more stable slope estimates than did the traditional fitting methods. Under certain conditions the procedure was able to identify the unstable performance. These results show that this interleaved tracking technique can produce more stable slope estimates, and can alert experimenters to drifting in subject performance.