Classification images for chromatic signal detection

The number and nature of the mechanisms for the detection of colored stimuli are still unclear. We use the paradigm of classification images to investigate the detection of a signal of homogeneous color added to a noisy texture. Both signal and noise colors were chosen from the isoluminant plane of the Derrington-Krauskopf-Lennie (DKL) color space. The signal consisted of a square of homogeneous color that was chosen from either cardinal or noncardinal directions of the DKL color space. The noisy texture consisted of small squares of varying colors that were chosen randomly across the isoluminant plane. Classification images reveal that (1) the cardinal axes play no specific role; (2) the widths of the tuning curves vary between 30 and 90 deg, consistent with the variation of tuning widths of neurons at early cortical stages; and (3) detection is not based on the whole region covered by the signal but is influenced mostly by a small spot around the fixation point.