Abstract
Psychophysical and physiological experiments have varied the angle of a (test or masking) stimulus in color space and taken a cosine pattern as evidence of linear cone combination in the studied mechanism or cell (e.g., the masking effect should vary with the cosine of the color angle between the mechanism direction and mask). However, this interpretation depends upon the color representation used: behavior of a linear mechanism will appear different when plotted in cone contrast and threshold-scaled cone spaces. The present experiments studied linearity of cone combination in mechanisms detecting S cone tests, using a method that is independent of color representation, and that allows a test for intrusion of multiple mechanisms by off-axis looking (which can mimic the effects of a nonlinearity). An S cone blob test was presented in noise that modulated, e.g., the L and M cones. Threshold for each compound noise (with both L and M cone components) was measured and compared to the threshold elevation for the components alone. Compound elevation in a linear mechanism can be predicted from the elevations from the components. For a wide range of conditions, for both increment and decrement S cone tests, the linear prediction fails.