A comparison of fixed-step-size and Bayesian staircases for sensory threshold estimation

Area of Science:

• Psychophysics
• Sensory perception
• Quantitative methods

Background:

• Fixed-step-size (FSS) and Bayesian staircases are standard for estimating sensory thresholds in two-alternative forced-choice (2AFC) tasks.
• Direct comparisons under identical conditions are scarce, limiting understanding of their relative efficacy.

Purpose of the Study:

• To directly compare the performance of optimized Bayesian staircases against optimized Fixed-step-size (FSS) staircases.
• To determine the optimal staircase method for experimental psychophysics based on estimate properties and cost.

Main Methods:

• A simulation study was performed to model staircase performance.
• An empirical test was conducted using optimized Bayesian and four variants of FSS staircases.
• Comparison focused on estimate bias, standard errors, and number of trials.

Main Results:

• Simulation: Both methods yielded dependable estimates with sufficient trials; Bayesian staircases showed lower standard deviations (higher efficiency).
• Empirical test: Neither procedure consistently converged; standard deviations did not decrease as expected; both showed similar efficiency.
• Discrepancies between simulation and empirical results were analyzed.

Conclusions:

• Simulations suggest Bayesian staircases are more efficient, but empirical data indicate similar performance.
• For practical psychophysics, using an FSS staircase with 18-30 reversals (1-up/3-down rule) is recommended.