Abstract
The small-sample nature of the typical psychophysical experiment presents us with the problem of finding valid, accurate statistical hypothesis testing methods. In many cases it can be difficult to obtain reliable error bounds on single threshold or slope estimates, for comparison across conditions or between subjects. Monte Carlo hypothesis testing methods, in particular the various developments of Efron's bootstrap technique for estimating confidence region boundaries, have enjoyed increasing popularity over the past fifteen years as computer speed has risen to meet their demands. I shall present a summary of some comparative simulations which looked at a number of variations on the bootstrap method, and showed differences in their stability. I shall also show how Monte Carlo simulations can help to assess the efficiency of one's algorithm for stimulus placement on the psychometric function.
