Task-dependent modulation of similarity judgments
Poster Presentation: Sunday, May 18, 2025, 2:45 – 6:45 pm, Banyan Breezeway
Session: Spatial Vision: Natural image statistics, texture
Schedule of Events | Search Abstracts | Symposia | Talk Sessions | Poster Sessions
Jonathan Victor1, Mary Conte1; 1Weill Cornell Medical College, New York, USA
Representations of visual information can often be described by perceptual spaces: constructs in which stimuli or features correspond to points, and their pairwise distances correspond to perceived dissimilarity. This information can be used in many ways – for example, discrimination, similarity, or grouping -- and correspondingly, a perceptual space’s geometry can change in a task-dependent fashion. The domain of visual textures is a model for probing these transformations. Using synthetic black-and-white textures (Victor & Conte 2015) and the paradigm of Waraich & Victor (2022) we recently showed that the transformation from the representational space for threshold judgments to suprathreshold similarity was characterized by selective stretching along some axes and curvature-producing nonlinear distortions (VSS 2024). Here, we examined a specific judgment, brightness comparison. On each of 1000 trials, subjects (N=5) ranked eight texture patches in order of similarity to a central reference texture. These judgments were used to infer a perceptual space via multidimensional scaling. The texture stimuli varied along many dimensions, including mean luminance and low- and high-order spatial correlations. In one experiment, subjects could judge similarity using any criterion. No further instructions were given. In a second experiment using the same stimuli, subjects were asked to judge similarity solely on the basis of the brightness of the central reference. Two novice subjects, unfamiliar with the texture stimuli and paradigm, carried out the brightness task first. Similar results were obtained from both the practiced and novice subjects. We expected that brightness judgments would yield a one-dimensional representational space, corresponding to luminance. Instead, the representational space remained high-dimensional. Compared to the representational space for suprathreshold similarity, the axis corresponding to luminance was elongated; other axes were selectively compressed; and curvature was less prominent. Thus, the perceptual space of similarity judgments can be selectively expanded or contracted by task.
Acknowledgements: NIH EY07977, Fred Plum Fellowship in Systems Neurology and Neuroscience