Abstract
Visual processing of images that contain broadband spatial content (e.g., a natural scene) is anisotropic (e.g., Essock, DeFord, Hansen, and Sinai, Vision Res., 2003) — horizontal content is perceived less well than content at other orientations and oblique content is seen best. That is, when typical scenes are being viewed, oblique content is most salient and has highest sensitivity, and salience and sensitivity is least at horizontal. This ‘horizontal effect’ has been linked to anisotropic low-level contrast gain control. Previously (Kim, Haun and Essock, VSS 2008), we have shown that separate gain control pools exist for low-speed (“sustained”) and high-speed (“transient”) mechanisms and that both show a horizontal effect. In the present experiments, we show that when overlay and surround suppression mechanisms can be isolated, both types of suppression yield horizontal effects. Using a grating target at two spatial frequencies (1cpd and 8cpd) and four orientations (0°, 45°, 90°, and 135° clockwise from vertical), contrast thresholds were measured for the gratings in the presence of either an overlaid (same-size) or annular surrounding patch of oriented 1/f noise at the same orientation as the target. Stimuli were presented at fixation or at one of several eccentric locations. Targets and masks were presented either with a slow Gaussian temporal waveform, or flickered at 16Hz. We found that at fixation, overlay masking dominates and yields a horizontal effect, while surround masking is relatively weak. Meanwhile, at increasing horizontal eccentricities, surround masking appears for the horizontal stimuli, is somewhat less for the vertical targets, and is barely present for oblique targets. Together with other findings, these results suggest that the horizontal effect is ubiquitous, and that wherever masking by 1/f noise is measurable, the anisotropy will appear.