Efficiencies for the statistics of size discrimination

Area of Science:

• Visual perception
• Psychophysics
• Computational neuroscience

Background:

• Human observers demonstrate consensus on estimating average orientation, limited by visual noise and inefficiency (√N objects used).
• Estimating average size in arrays of circles is controversial regarding observer efficiency.
• Comparing size and orientation estimation against an ideal observer is crucial for quantitative statistical efficiency evaluation.

Purpose of the Study:

• To measure visual noise limiting size estimates between two circles.
• To assess human observer efficiency in estimating average size from arrays of circles.
• To measure just-noticeable differences in size variance and compare efficiencies.

Main Methods:

• Measured Weber fractions for size discrimination (0.07–0.14 for 84% correct 2-alternative forced choice).
• Exaggerated visual noise by varying circle diameters in arrays (1, 2, 4, 8 circles).
• Measured discrimination of mean sizes and size variances.

Main Results:

• Size estimation efficiency exceeded 25%, approaching 100% in some cases.
• Variance discrimination efficiencies ranged from 62% to 75%.
• Observers were more precise in size variance discrimination than mean size discrimination.

Conclusions:

• Human visual system adds Gaussian noise to logarithmically transduced circle diameters.
• Observer efficiency for mean size estimation is significantly higher than previously thought.
• Evidence suggests a late-stage noise source limits mean size discrimination but not variance discrimination.