Uncovering the geometry of color space with magnetoencephalography (MEG)
Rosenthal, I.; Singh, S.; Hermann, K.; Pantazis, D.; Conway, B. R.
Show abstract
The geometry that describes the relationship among colors is unsettled despite centuries of study. Here we present a new approach, using multivariate analyses of direct measurements of brain activity obtained with magnetoencephalography to reverse-engineer the geometry of the neural representation of color space. The analyses depend upon determining similarity relationships among the neural responses to different colors and assessing how these relationships change in time. To evaluate the approach, we relate patterns of neural activity to universal patterns in color naming. Control experiments showed that responses to color words could not decode activity elicited by color stimuli. The results suggest that three patterns of color naming can be accounted for by decoding the similarity relationships in the neural representation of color: the association of warm colors such as reds and oranges with "light" and cool colors such as blues and greens with "dark"; the greater precision among all languages in naming warm colors compared to cool colors; and the preeminence of red.
Matching journals
The top 5 journals account for 50% of the predicted probability mass.
Similar papers in this journal
- Automatic and fast encoding of representational uncertainty underlies probability distortion 96%
- Stimulus-modality independent activation of the brain network for language 96%
- Deep neural networks and visuo-semantic models explain complementary components of human ventral-stream representational dynamics 95%
Similar papers in this journal
Similar papers in this journal
- Efficient prestimulus network integration of FFA biases face perception during binocular rivalry 95%
- Linking the brain with behaviour: the neural dynamics of success and failure in goal-directed behaviour 95%
- Comparing Neural Correlates of Memory Encoding and Maintenance for Foveal and Peripheral Stimuli 94%
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.