Feature-based attention modulates population spatial frequency tuning

Poster Presentation 43.338: Monday, May 22, 2023, 8:30 am – 12:30 pm, Banyan Breezeway
Session: Attention: Features

Luis D. Ramirez1 (), Feiyi Wang2, Sam Ling1; 1Boston University, 2Tufts University

To enhance perception, it has been proposed that attention alters the receptive field properties of neurons tuned to the attended feature — effectively transforming visual processing downstream. Does attention toward a specific spatial frequency then alter spatial frequency processing in early visual cortex? Here, we used functional magnetic resonance imaging to measure changes in population spatial frequency tuning (pSFT; see Aghajari, Vinke, & Ling, 2020), while subjects attended to specific spatial frequencies. We manipulated feature-based attention to spatial frequency by cueing subjects to perform a letter detection task on one of two super-imposed letter streams, positioned in one visual hemifield (eccentricity = 3.5°; diameter = 3°). Each letter stream was bandpass filtered to contain only low (0.5 cpd) or high (2 cpd) spatial frequency content (filter width = 0.1). In the opposite visual hemifield, we presented a probe stimulus, which swept through a range of bandpass filtered Gaussian white noise stimuli (40 logarithmically spaced frequencies between 0.1 and 12 cpd; filter width = 0.1; 80% contrast) to acquire voxel-wise estimates of pSFT across V1–V3. While this probe stimulus was not spatially attended, we leveraged the known spatial spread of feature-based attention to independently interrogate its influence on spatial frequency tuning. With the acquired pSFT, we tested various models for how feature-based attention might modulate tuning: a change in the peak, the bandwidth, or the amplitude (and combinations of these mechanisms). Preliminary results suggest that feature-based attention uniquely alters pSFT across the visual field and early visual areas, including an attentionally-driven shift in peak spatial frequency tuning preferences.

Acknowledgements: Funded by National Institutes of Health Grants: R01EY028163 to S. Ling and F99NS124144 to L. D. Ramirez