Gamma oscillations across recording scales show a preference for saturated long-wavelength (reddish) hues in the primate visual cortex

Abstract

Stimulus-induced narrowband gamma oscillations (30-70 Hz) arise due to excitatory-inhibitory interactions and have been proposed to carry feedforward prediction errors during visual processing. However, the dependence of gamma on stimulus color is not well characterized, with some studies showing a preference for saturated long-wavelength (reddish) hues in invasive recordings from the primary visual cortex (V1), while others showing no preference for red in luminance and cone-contrast balanced colors spanning the Derrington-Krauskopf-Lennie (DKL) isoluminant plane in non-invasive magnetoencephalography (MEG) recordings. To address these discrepancies, we simultaneously recorded local field potentials (LFPs) from V1 (n=2 monkeys) along with scalp electroencephalography (EEG), while presenting luminance-matched hues spanning the entire permissible range of colors. We found that saturated reddish hues produced the strongest gamma in both LFP and EEG in both monkeys. However, gamma was reduced when the colors were desaturated, so that selectivity for reddish hues was diminished or absent on the DKL space, as shown previously. Interestingly, selectivity was further reduced in EEG compared to LFP. Simultaneous recordings from an intermediate visual area (V4, n=1) revealed weak color-induced gamma, partially explaining the weakening of hue selectivity in macro-signals. Gamma power increased with increasing activation along the red-green (L-M) cardinal axis but was virtually absent along the blue-yellow (S-(L+M)) axis, suggesting that gamma could be a reflection of the L-M cone-contrast mechanism. These results comprehensively resolve the earlier discrepancy and shed light about the neural mechanisms underlying gamma generation in primate V1. Significance StatementGamma-band brain rhythms are markers of cortical computation and disease, yet their dependence on color has remained controversial because studies using different recording methods and stimulus color spaces yielded conflicting results. By densely sampling colors across the entire monitor gamut in an isoluminant space and mapping brain rhythms across multiple neural scales in primates, we show that primary visual cortex generates exceptionally strong gamma for saturated reddish hues. Crucially, this effect is driven by contrast in specific cone pathways (specifically, L-M cone contrast), and the selectivity diminishes in desaturated color spaces, downstream visual areas and scalp recordings. This explains why the "red bias" observed invasively often vanishes in non-invasive recordings and demonstrates that color-induced gamma reflects pathway-specific circuit engagement.