Salience from feature contrast: temporal properties of saliency mechanisms

Abstract

Single cell recordings in area V1 of the macaque monkey had suggested that saliency effects from orientation contrast might be delayed compared to the representation of other stimulus properties. This conjecture was tested in three series of experiments on regular line patterns. Experiment 1 investigated the time courses of saliency effects evoked either by the onset of a single line or by a target that popped out from orientation contrast. Saliency effects from orientation contrast developed later than saliency effects related to stimulus onset. Experiment 2 measured the detectability of such targets in brief presentations. As expected, single line targets were detected at shorter presentation times than popout targets with orientation contrast. Experiment 3 finally investigated the temporal resolution of saliency effects from feature contrast in different dimensions. Line arrays with a popout target (e.g. an orthogonal line) were alternated with complementary line arrays in which the target and the non-target features were exchanged (e.g. all lines were orthogonal to those in the previous pattern). Thus, although feature contrast was present in every single stimulus display, saliency effects could only develop when alternation rates were slow enough to be resolved by the underlying saliency mechanisms. Feature flicker of this sort was tested in orientation, motion (direction), color and luminance. Saliency mechanisms encoding orientation contrast were slower than those encoding differences in luminance or color; motion contrast produced intermediate results that also differed between subjects.