Keynotes | K&Ts | GACs | Talks | Posters | Search

Poster E in Poster Session E: Thursday, August 6, 10:30 am – 12:15 pm, Kimmel Center, Shorin & Rosenthal Rooms

Graded Representation of Feature Salience in Neural Feature Dimension Maps

Daniel Thayer¹, Emily Machniak¹, Christina Melikian¹, Thomas Christopher Sprague¹; ¹University of California, Santa Barbara

Presenter: Daniel Thayer

Visual attention is automatically directed to salient stimuli, such as a suddenly moving dog. This is reflected within 'feature dimension maps', which prioritize specific locations based on factors such as feature-specific image salience (motion map would prioritize moving dog). High-priority locations are more likely to be selected by attention for further processing than low-priority locations. Even though previous work identified retinotopic regions of cortex that preferentially respond to feature-defined salient stimuli (color or motion), consistent with neural correlates of feature dimension maps, it is unclear if these regions compute the relative level of salience across contrasts. Here, we tested if neural feature dimension maps, as measured using image reconstruction techniques with fMRI, contain graded salience representations by presenting displays where a location was made salient by adjusting either the color or motion stimulus contrast. Neural feature dimension maps tracked relative stimulus salience in a manner consistent with feature preferences. Furthermore, salient motion and color had different neural response profiles, where color contrast was represented monotonically and motion was represented non-monotonically. These results indicate that computations of salience occur within specialized cortical areas, informing models of visual attention by characterizing where feature salience is neurally represented.

Topic Area: Computational Models of Vision & Visual Cortex