Slow-varying normalization explains diverse temporal frequency masking interactions across scales in the macaque visual cortex

Neurons in the primary visual cortex (V1) respond non-linearly with the presentation of multiple stimuli, which has been explained by a normalization model where the excitatory drive is divided by the summed activity of a large neuronal population. While recent studies have suggested that normalization could be time and frequency-dependent, neural mechanisms underlying this dependence remain unknown. Steady-state visually evoked potentials (SSVEPs), which are produced by presenting flickering or counterphasing visual stimuli, serve as a robust tool to probe these underlying mechanisms by leveraging frequency-specific tagging of concurrently presented stimuli. We presented two overlapping counterphasing grating stimuli (plaids), either parallelly or orthogonally, at multiple contrasts and temporal frequencies and recorded spikes, local field potential, and electrocorticogram from V1 of bonnet macaques while they passively fixated. We also recorded electroencephalogram (EEG) activity. The resulting SSVEPs exhibited complicated dynamics - with "low-pass" and "band-pass" suppression profiles for orthogonal and parallel plaids, respectively. Importantly, these dynamics were conserved across scales - from spiking activity to EEG. Surprisingly, adding a simple low-pass filter in the normalization signal sufficiently explained these diverse effects. Our results present a simple mechanism to explain the spectro-temporal dynamics of normalization. These insights may also aid in designing and interpreting SSVEP-based cognitive and EEG-based brain-computer interfacing (EEG-BCI) studies.