Efficient estimation of stereo thresholds: What slope should be assumed for the psychometric function?

Area of Science:

• Psychophysics
• Computational Neuroscience
• Vision Science

Background:

• Bayesian staircases are crucial for estimating detection thresholds in psychophysics.
• Accurate parameter selection for the psychometric function is vital for precise threshold estimation.
• The slope (or spread) of the psychometric function is often fixed, risking bias if incorrect.

Purpose of the Study:

• To determine the optimal psychometric function slope for minimizing bias and maximizing precision in stereoacuity measurement using Bayesian staircases.
• To empirically measure the spread of disparity psychometric functions in human observers.
• To provide evidence-based recommendations for selecting slope parameters in Bayesian staircase methods.

Main Methods:

• Conducted 2-alternative forced choice (2AFC) and 4AFC disparity detection stereo experiments.
• Measured the spread of the disparity psychometric function assuming a Logistic function.
• Performed simulations using empirical data to identify optimal spread values.
• Compared the performance of a Bayesian procedure (ZEST with optimal σ) against other Bayesian methods.

Main Results:

• Observed a wide range of disparity psychometric function spreads (0.03 to 3.5 log10 arcsec) with minimal age-related changes.
• Simulations indicated optimal assumed spread values between the 60-80th percentiles of the population distribution.
• Recommended stereo threshold spread values: σ ≈ 1.7 log10 arcsec for 2AFC and σ ≈ 1.5 log10 arcsec for 4AFC.
• ZEST with optimal σ demonstrated the lowest threshold bias and highest precision among tested methods.

Conclusions:

• Selecting an optimal, empirically-derived spread parameter significantly enhances the accuracy of Bayesian staircase threshold estimations.
• The recommended spread values offer improved performance for stereoacuity measurements and can be adapted for other threshold types.
• ZEST utilizing an optimized spread parameter provides a robust and precise method for psychophysical threshold determination.