Intrinsic Organization of Contrast Sensitivity in Human Vision

Natural vision operates over an enormous range of luminance while preserving sensitivity to both fine and coarse spatial structure. As luminance increases, the peak of the contrast sensitivity function shifts toward higher spatial frequencies. Existing accounts explain how luminance rescales the amplitude of neural responses through gain control, but not why preferred spatial frequency should reorganize systematically with luminance. Here we propose that recurrent excitatory-inhibitory interactions generate an intrinsic spatial frequency, set by the coupling structure of the circuit, that shapes the network's resonance and thereby predicts changes in preferred spatial frequency. Changes in mean luminance shift this intrinsic spatial frequency by altering the effective balance between excitation and inhibition, thereby moving the network between dynamical regimes rather than simply rescaling its responses. To test this framework, we measured human contrast sensitivity across a finely sampled range of luminance. Preferred spatial frequency followed the predicted course: approximately constant at low luminance, then rising sharply over a narrow range in an S-shaped transition. These results support a circuit-based account in which changes in cortical state reorganize spatial selectivity, linking luminance-dependent perception to the dynamics of cortical computation.