Farhan Abdul Vaheed¹, Allison B. Sekuler^1,2,3, Patrick J. Bennett¹; ¹Department of Psychology, Neuroscience, and Behaviour, McMaster University, ²Rotman Research Institute, Baycrest Academy for Research and Education, Toronto, Canada, ³Department of Psychology, University of Toronto

How do we discriminate visual patterns? One view is that discrimination is based on changes in the distribution of responses across multiple local filters tuned to different spatial frequencies, orientations, and directions of motion. However, some evidence suggests that spatial-interval (Morgan & Ward, JOSA, 1985) and relative phase (Badcock, Vision Res, 1984) discrimination utilizes representations that explicitly represent the spatial characteristics of visual stimuli. Here, we examine if observers use changes in the luminance spatial profile to discriminate patterns in a putative contrast discrimination task. The stimuli were f + 2f compound gratings. The contrast of f and 2f were 0.1 and 0.05, respectively, and a 3IFC task was used to measure discrimination thresholds for a contrast increment added to 2f. Thresholds were measured in a fixed baseline condition and in conditions in which we randomized stimulus contrast (±0.3 log units), spatial frequency (±0.3 log units), or orientation (±15°). In the compound random conditions, randomization was applied to the compound stimulus, where both components changed by equal amounts, and therefore the stimulus luminance profile remained unchanged. In the component random conditions, randomization was applied to only one grating component, and therefore, the spatial profile was altered. We found that thresholds in the component random conditions, but not the compound random conditions, were significantly greater than the threshold in the fixed, baseline condition. In other words, discrimination thresholds were nearly constant in conditions where observers could base discrimination on changes in the luminance spatial profile, but were much higher when that cue was not available. These findings suggest that if discriminations are mediated solely by spatially localized filters, then those filters must encode aspects of local waveform shape, and that the integration of information across channels is disrupted when the pattern’s luminance profile is changed.

Acknowledgements: NSERC