Abstract
Visual analysis is optimized through the selective sampling of the most salient regions in the scene. Several factors may contribute to the definition of local salience. Here we chose a specific set of local features, predicted by a constrained maximum-entropy model to be optimal information-carriers, as candidate “salient features”. Psychophysical and eye movements studies, in which participants were instructed to choose the most salient stimulus, have shown that, in fast vision, these features are perceived as more salient than others, even when shown in isolation. Here, we aimed to show that these features “automatically” attract attention and eye movements in implicit tasks. In a spatial orienting covert-attention task, we presented a tilted gabor on the left or on the right of the fixation point (5º) with different contrasts, preceded by two brief peripheral cues (26 ms), one more salient than the other according to the model. The salient cue is presented on the same side of the gabor in valid trials, with 50% or 80% cue validity. Subjects had to identify the orientation of the gabor as a function of gabor contrast while keeping fixation. In the overt attention task, participants had to perform a saccade towards a circular place holder (target), that could be presented at the salient cue location with 50% or 80% cue validity. Results are coherent in both tasks, showing lower contrast thresholds and saccade latencies in valid trials and higher thresholds and saccade latencies in invalid trials, independently of cue validity, compared to baseline values obtained with equally salient cues. Similar effects were found in a control condition where cues saliency was manipulated through their relative luminance. These findings suggest that in fast vision covert and overt attention are automatically attracted by the saliency provided by the optimal information-carriers features predicted by the reference model.