Abstract
Abstract: :
Purpose: Last year we reported an "horizontal effect" in which detection of added oriented structure in either broadband noise patterns or natural scenes was poorest for horizontally oriented structure, and best for obliquely oriented structure (DeFord et al., ARVO 2001). In the present study we further investigate this anisotropy by: (1) measuring the salience of suprathreshold oriented structure embedded in noise, and (2) manipulating the content of the broadband noise to determine the properties of the noise that lead to an horizontal effect rather than the more typical oblique effect obtained with narrowband stimuli. Methods: Oriented noise patterns were constructed by incrementing the amplitude spectra of isotropic noise patterns (i.e., random phase) within a given orientation band. The amplitude spectra tested included the 1/f spectrum that characterizes natural scenes and manipulations of slope, orientation bandwidth, and frequency bandwidth. A standard matching paradigm was used to assess the perceived strength of the added oriented structure in a test pattern by having the observers adjust the amplitude of the oriented structure in a comparison pattern to make a match. Results: Manipulations in the slope and bandwidth of the noise pattern's amplitude spectrum resulted in varying magnitudes of an horizontal effect, and in some cases resulted in an "inverse oblique effect" (i.e., the salience of vertical content was diminished like that of horizontal content). Conclusion: These results further support the conclusion that the visual system automatically adjusts the perceptual strength of oriented content when viewing broadband noise and natural scenes. However, the anisotropic nature of this "gain" adjustment is dependent on the broadband content present in a scene leading to an horizontal perceptual bias in some conditions, and to an inverse oblique effect in other conditions.
Keywords: 586 spatial vision • 584 scene perception