Fig. 1: Decoding luminance polarity and hue from MEG data: approach.

a Decoding luminance polarity. Participants were shown four hues that varied in luminance polarity (light/dark). Classifiers were trained (solid arrows) and tested (dashed arrows) using stimuli of the same hue (identity problems) or different hue (generalization problems) to determine the extent to which the MEG response to a given color is informative of the luminance polarity carried by the same hue (4 identity problems) or by other hues (12 generalizing-across-hue problems). Each binary classifier was trained to distinguish whether a light or dark stimulus had been presented, given patterns of magnetoencephalography (MEG) sensor activations. For the four identity problems, classifiers were trained and tested on brain responses to the same hue (all four identity problems are illustrated in the graphic). For the generalization problems, classifiers were trained and tested on brain responses to different hues (half of the 12 generalizing-across-hue problems are illustrated). In the graphic, solid lines indicate comparisons used for training and dashed lines indicate comparisons used for testing; line shading distinguishes different problems. b Decoding hue. Format as in panel a. For stimuli of a given luminance polarity (e.g., dark), a binary classifier was trained to determine which of two hues (e.g., pink or orange) had been presented. The classifiers were then tested, again on held-out trials in which the luminance polarity (e.g., dark) was the same as at train time (12 identity problems) or in which the luminance polarity was opposite (e.g., light), requiring generalization of hue across luminance polarity (12 generalizing-across-luminance-polarity problems).