A comparison of two-, three-, and four-alternative forced-choice staircase procedures

Threshold estimates from multiple-interval forced-choice staircase procedures were studied using computer simulations. Two frequently cited psychometric function shapes, those for the signal known exactly and for high uncertainty, governed the hypothetical subject's responses in the simulations. Threshold estimates based on 100 trials were obtained for both psychometric functions, for three different stimulus change values, and for decision rules that estimated 70.7% and 79.4% correct levels. A total of 1000 replications in each condition were run. Each threshold estimate was calculated by averaging the stimulus levels at which a reversal in stimulus direction change occurred. For many of the simulated conditions, three-alternative forced choice was the most efficient procedure (i.e., the variability of the estimates per unit of collection time was smallest). The validity of the simulations was tested using human performance data. [Work supported by NINCDS and IBM.]