Abstract
Visual features such as luminance and orientation are known to influence oculomotor behavior. However, as suggested in models of saccade generation, they may intervene with different time courses, through respectively direct (retinal) vs. indirect (retino-cortical -V1) projections to the Superior Colliculus. To test this hypothesis, we compared the time required to initiate a saccade toward a peripheral visual target that was defined either by its average luminance or its orientation relative to a textured background. In a first experiment, we used a forced-choice saccade task (Kirchner & Thorpe, 2006), in which the target appeared randomly to the left or to the right of an initial fixation cross, and at a variable eccentricity. The target was a vertical array of Gabor patches whose luminance and orientation relative to the background (a grid of 25°-tilted Gabor patches) were varied randomly (method of constant stimuli). This allowed us to estimate for each participant, the luminance/orientation thresholds that yielded a saccade in the correct direction in 98% of the cases, and hence determine the levels that made luminance and orientation targets equally salient. In the second experiment, we then measured saccadic reaction times (SRT) for these salience-matched, luminance and orientation targets. We observed that SRT were longer for orientation- compared to luminance-defined targets. Irrespective of target eccentricity, individual SRT distributions were shifted by about 20 ms on average, towards longer latencies for orientation- compared to luminance-defined targets, while saccade accuracy remained unaffected. Thus, orientation intervenes later than luminance in determining when the eyes move, probably as a result of the different neural substrates involved in the extraction of these two visual features, and the slightly longer route to the SC for orientation compared to luminance. Implications for models of eye guidance in natural scenes, where luminance and orientation presumably contribute equally will be discussed.
Meeting abstract presented at VSS 2014