Guidance of Gaze Based on Color Saliency in Monkeys with Unilateral Lesion of Primary Visual Cortex

In the accompanying paper (Itti et.al.), we investigated residual visually-guided behavior in monkeys after unilateral ablation of primary visual cortex (V1), to unravel the contributions of V1 to salience computation. We analyzed eye movements of monkeys watching video stimuli and a computational model of saliency-based, bottom-up attention quantified the monkeys' propensity to attend to salient targets. All monkeys were attracted towards salient stimuli, significantly above chance, for saccades directed both into normal and affected hemifields. We also quantified the contribution of visual attributes (intensity, color, motion and so on) to the saliency-based eye movements and obtained evidence that the monkeys' guidance of gaze was influenced by color saliency. Here we directly examined residual visuomotor processing based on color saliency with color discrimination tasks. In two monkeys after unilateral ablation of V1, the isoluminant, chromatic stimuli was presented in one of the two positions in their affected hemifield. The monkeys were rewarded by making saccade to the target. The CRT monitor (Mitsubishi DZ21) was used for stimulus presentation and was calibrated with a colorimeter (PR650). The stimuli were defined by the DKL color space, that is, the luminance axis, the L-M axis and the S-(L+M) axis. In both monkeys, the correct ratio was significantly above chance for stimuli with the L-M component and the S-(L+M) component. Control experiments were done to exclude the possibility that a small luminance difference from background may contribute to the above-chance performance. When a small positive or negative luminance difference (<5%) was added to the chromatic stimuli, the correct ratio was not decreased. On the other hand, the correct ratio was near the chance level when the achromatic stimuli with the same luminance difference were used. Our results suggest that unilateral ablation of V1 does not abolish the computation of color saliency.