Letting external noise rather than internal noise limit discrimination performance allows information to be extracted about the observer's stimulus classification rule. A perceptual classification image is the correlation over trials between the noise amplitude at a spatial location and the observer's responses. If, for example, the observer followed the rule of the ideal observer, the response correlation image would be an estimate of the ideal observer filter, the difference between the two unmasked images being discriminated.
Perceptual classification images were estimated for a vernier discrimination task. The display screen had 48 pixels per degree horizontally and vertically. The no-offset image had a dark horizontal line of 4 pixels, a 1 pixel space, and 4 more dark pixels. Classification images were based on 1600 discrimination trials with the line contrast adjusted to keep the error rate near 25 percent. In the offset image, the second line was one pixel higher. Unlike the ideal observer filter (a horizontal dipole), the observer perceptual classification images are strongly oriented. Fourier transforms of the classification images had a peak amplitude near one cycle per degree and an orientation near 25 degrees. The spatial spread is much more than image blur predicts, and probably indicates the spatial position uncertainty in the task.
Problems
1) Observer can change rules with changes in conditions.
2) Levels of performance are dependent on factors other than the classification rule.
Luminance: 26.5 cd/m^2
Area: 13.5 deg x 10.1 deg (640 x 480 pixels)
Uniform Random Noise
Peak Contrast: + or - 0.25
Area: 2.7 deg x 2.7 deg (128 x 128 pixels)
Vernier Lines
Width: 5.06 min (4 pixels)
Height: 1.26 min (1 pixel)
Gap: 1.26 min (1 pixel)
Offset: 1.26 min (1 pixel)
Contrast: -0.23 (ND) -0.38 to -0.44 (AA)
(adjusted for $approx$ 75% correct)
Duration: 1.0 sec (ND) 0.5 sec (AA)
Background
Luminance: 41.1 cd/m^2
Area: 6.4 deg x 4.8 deg (800 x 600 pixels)
Uniform Random Noise
Peak Contrast: + or - 0.125
Area: 1.0 deg x 1.0 deg (128 x 128 pixels)
Vernier Lines
Width: 2.4 min (4 pixels)
Height: 0.48 min (1 pixel)
Gap: 1.44 min (3 pixels)
Offset: 0.48 min (1 pixel)
Contrast: -0.74
Duration: 0.5, 1.0 sec
Noise amplitudes some distance from the lines in both horizontal and vertical directions made significant contributions to the responses.
Unlike the ideal observer, the observers used information near the fixed line almost as much as that near the variable line.
The classification images are not consistent with the discrimination being based on the output of a single Gabor-like filter.