Temporal Evolution of Color Representations Measured with Magnetoencephalography Reveals a "Coarse to Fine" Dynamic.

Abstract

Color perception is based on the differential spectral responses of the L, M, and S-cones and subsequent subcortical and cortical computations and may include the influence of higher-order factors such as language. Although the early subcortical stages of color vision are well characterized, the organization of cortical representations of color remain elusive, despite numerous models based on discrimination thresholds, appearance, and categorization. An underexplored aspect of cortical color representations is their dynamic evolution. Here, we compare the evolution of three different color representations over time using magnetoencephalography. We measured neural responses to 14 hues at each of three achromatic offsets (increment, isoluminant, and decrement) while participants attended either to the exact color of the stimulus or its color category. We used a series of classification analyses, combined with multidimensional scaling and representational similarity analysis, to ask how cortical representations of color unfold over time from stimulus onset. We compared the performance of "higher order" models based on hue and color category with a model based simply on stimulus cone contrast and found that all models had significant correlations with the data. However, the unique variance accounted for by each model revealed a dynamic change in hue responses over time, which was consistent with a "coarse to fine" transition from a broad clustering into categorical groups to a finer within-category representation. Notably, these dynamics were replicated across data sets from both tasks, suggesting they reflect a robust reorganization of cortical hue responses over time.