Modeling Visual Representations During Manipulations of Covert and Overt Attention
vRF model fits across the cortex (A) Location (polar angle; see inset) of each voxel’s preferred position based on vRF model fitting (only voxels with 10% variance explained or greater colored). Example participant (sub003) shown; posterior view. (B) vRF models for a subset of voxels shown in screen coordinates (+ indicates fixation point; color matches (A)). (C) Raw fMRI signal from each voxel in (B), sorted by polar angle. Comparing top to bottom, the peaks of the fMRI response shift earlier or later in time, depending on how each voxel responds to the stimulus sequence.
This project aims to leverage recently-developed computational neuroimaging methods whereby mathematical models are used to predict brain responses according to known sensory input and cognitive task conditions to characterize and quantify limitations inherent in visual information processing within the visual and parietal cortex of healthy human participants (civilians). Rather than identifying which brain regions ‘light up’ or which are more or less strongly activated between task conditions, this approach requires building a comprehensive and predictive understanding of neural information processing, and probing this understanding with novel ‘challenges’—new types of visual stimuli or task conditions and their combinations.
In this project, we will extend the model-based fMRI approach developed in our work thus far to test how manipulations of visual stimulus and task parameters impact neural representations across several new experimental scenarios. This work is innovative because it involves building and testing competing predictive models of visual responses across well-controlled task manipulations. If these predictive models are efficacious and are demonstrated to generalize across experimental manipulations we will have discovered fundamental computational principles used by neural systems to optimize information processing. Additionally, because we will quantitatively formalize each model’s expected changes to visual processing under different cognitive and stimulus manipulations, these models can be used in other experiments and by other research groups to probe additional experimental scenarios not envisioned here.
