Attentional Modulation of Spatial Frequency Tuning: Links to Global–Local Processing and Potential Remediation of Frequency Biases

Selective attention is a fundamental cognitive process that enables the filtering of irrelevant sensory information in favor of task-relevant inputs. Early visual areas such as the primary visual cortex (V1), V2, and V3 are involved in encoding sensory input in a feedforward manner, but these regions are also dynamically modulated by top-down signals, including signals that mediate visual attention (Carrasco, 2011; Liu, 2019). These attentional signals, which are largely thought to originate in frontoparietal regions including the frontal eye fields and intraparietal sulcus, can enhance neural responses in early visual cortex by increasing firing rates evoked by relevant stimuli (response gain), improving the signal-to-noise ratio, and sharpening or shifting tuning curves (e.g., for contrast or spatial location; Carrasco, 2011; Liu, 2019). Spatial frequency encoding is a fundamental function of early visual cortex, which supports the extraction of both fine and coarse visual detail. In V1, neurons are arranged retinotopically, with systematic biases in spatial frequency tuning superimposed: foveal regions are more sensitive to high spatial frequencies (e.g., fine edges and textures) and peripheral regions to low spatial frequencies (e.g., coarse shapes and contours); similar eccentricity-linked gradients occur in V2 and V3 (Aghajari et al., 2020). These patterns reflect cortical magnification—the disproportionate allocation of cortical territory to central vision—which enables fine detail to be resolved in the fovea while coarser detail dominates in the periphery. How selective attention modulates these population-level spatial frequency representations and whether such modulation respects or reshapes these underlying biases are less clear. Ramirez et al. (2025) addressed this using fMRI to measure visual cortical activity while participants attended to either low- or high-spatial frequency letter streams (0.5 vs 2.0 cycles/degree) in one hemifield as they simultaneously view spatial frequency-filtered noise in the opposite, unattended hemifield. Because feature-based attention is known to spread globally … Correspondence should be addressed to Hayden J. Peel at hayden.peel{at}camh.ca.