Due to these shortcomings, researchers have found numerous ways to improve the experimental methods over the past few decades. A large number of these are surveyed and compared in [331], and fall under the heading of adaptive psychophysical methods. Most improved methods perform staircase procedures, in which the stimulus magnitude starts off with an easy case for the subject and is gradually decreased (or increased if the reference magnitude is larger) until the subject makes a mistake [91]. At this point, the direction is reversed and the steps are increased until another mistake is made. The process of making a mistake and changing directions continues until the subject makes many mistakes in a short number of iterations. The step size must be carefully chosen, and could even be reduced gradually during the experiment. The direction (increase or decrease) could alternatively be decided using Bayesian or maximum-likelihood procedures that provide an estimate for the threshold as the data are collected in each iteration [114,156,350]. These methods generally fall under the heading of the stochastic approximation method [270].
Steven M LaValle 2020-11-11