Abstract
Electroencephalography (EEG) has been used for decades to study the dynamics of color processing. Chromatic visual evoked potentials (cVEPs) to stimuli preferentially activating the cardinal cone-opponent mechanisms have distinct latencies and amplitudes. However, the degree to which different chromatic responses are separable from one another, and from the response to a chromatically neutral point, remains largely undetermined. The current study used multivariate pattern analysis (MVPA) to investigate this issue.
VEPs for three subjects were evoked by Gabor stimuli that varied in chromatic contrast from a white point in eight directions in cone-opponent space. Each stimulus was presented in onset-offset format 120 times, and EEG data were recorded from a single occipital electrode. A linear discriminant analysis algorithm was trained and tested using leave-one-trial-out cross-validation, and permutation tests were used to determine statistical significance of classification. The voltage at each time point was used as a feature, with classification based on ~760 features.
With the exception of one subject for whom the –S direction could not be accurately classified, cVEPs were able to accurately distinguished from white point VEPs in all eight chromatic directions for all three subjects. Classification between different chromatic directions was more variable. Most pairs of axes could be distinguished from one another, and failures in classification were usually between opposite directions (e.g. +L and −L). Thus the time course of activity recorded from one electrode can be used not only to classify a stimulus as colored versus achromatic, but also to distinguish between different colors.