The paper by Wichmann and Hill (2001) discusses the estimation of variability in fitted parameters and derived quantities, such as thresholds and slopes, of psychometric functions. They advocate the use of the bootstrap method, a Monte Carlo resampling technique, to estimate variability accurately, especially given the small size of typical psychophysical data sets. The authors outline the basic bootstrap procedure, favoring the parametric bootstrap over the nonparametric one. They describe how to test the validity of the bootstrap bridging assumption, which is crucial for the accuracy of the bootstrap estimates. The choice of sampling scheme significantly affects the reliability of bootstrap confidence intervals, and the authors provide recommendations for efficient sampling. They also address the influence of the distribution function on the variability estimates, noting that while the choice of distribution function generally has a minor impact, it can be significant for lower response thresholds. Finally, they introduce improved confidence intervals (bias-corrected and accelerated) that enhance the accuracy of the bootstrap estimates.The paper by Wichmann and Hill (2001) discusses the estimation of variability in fitted parameters and derived quantities, such as thresholds and slopes, of psychometric functions. They advocate the use of the bootstrap method, a Monte Carlo resampling technique, to estimate variability accurately, especially given the small size of typical psychophysical data sets. The authors outline the basic bootstrap procedure, favoring the parametric bootstrap over the nonparametric one. They describe how to test the validity of the bootstrap bridging assumption, which is crucial for the accuracy of the bootstrap estimates. The choice of sampling scheme significantly affects the reliability of bootstrap confidence intervals, and the authors provide recommendations for efficient sampling. They also address the influence of the distribution function on the variability estimates, noting that while the choice of distribution function generally has a minor impact, it can be significant for lower response thresholds. Finally, they introduce improved confidence intervals (bias-corrected and accelerated) that enhance the accuracy of the bootstrap estimates.