Abstract
Attention is thought to bias competitive and cooperative interactions in a visual processing hierarchy. Coherent pattern motion (Adelson & Movshon, '82) constitutes a good test for this hypothesis, as it involves at least two hierarchical levels. Here we are interested in how attention alters the respective representations of a single pattern and of two competing patterns of motion (see also Braun, Pastukhov, VSS'04). We report two dual-task experiments with an attention-demanding task near fixation (5 shapes rotating anti- or clock-wise) and a circular array of coherent pattern motion in the near periphery (72 log-Gabor wavelets, 3 array diameter, 4 eccentricity). In Exp. 1, observers discriminate the overall direction/orientation ? of three alternative types of arrays: (i) randomly oriented wavelets with individually adjusted speed vi=vmax cos(?i−?) (coherent motion), (ii) parallel wavelets with uniform speed (parallel motion), or (iii) parallel wavelets with zero speed (static control). Comparing fully and poorly attended arrays (i.e., single- and dual-task condition), we find a ∼20% increase for coherent motion, but a ∼150% increase for parallel motion and static control. This shows dramatically smaller attention effects for moving arrays in the absence of static orientation cues. In Exp. 2, observers report (“Sticking or sliding?”) the perceived coherence of two superimposed parallel motion arrays. Keeping one array's spatial frequency fixed, we systematically vary that of the other (contrast also remains fixed). Compared to poor attention, full attention shifts the point where ‘sticking’ and ‘sliding’ are equally likely along the spatial frequency axis, such as to enlarge the ‘sliding’ regime. The size of this shift depends on spatial frequency. Thus, attention substantially changes the interaction between two moving patterns that are transparently superimposed. This opens a new window on attentional changes in the processing of visual motion.
Support: EU IST-2001-38099