Fig. 6. Cortical mechanism achieving processing of object surface and edge.

a Illustration of two possible models for the generation of surface responses in the output layer. Model A (upper panel), neuronal responses in the output layer are generated by pooling excitatory projections from the V1 input layer. Model B (lower panel), neuronal responses in the output layer are generated by combining the excitations and nonlocal inhibitions in the output layer. b Performance comparison for explaining the neuronal response in the output layer between model A (x-axis) and model B (y-axis). n = 45 sites from output layers 2, 3 and 4B (same for c–h). c Performance improvement from model A to model B. ***p < 0.001; n.s.: p = 0.471, two-sided paired t-test. Bars (in c–g) present mean of corresponding values across sites (±s.e.m.), with individual data superimposed (n = 45). d, e Predicted surface suppression (d) and edge suppression (e) from the two models. ***p < 0.001; n.s.: p = 0.2806 (in d) and p = 0.0973 (in e); two-sided paired t-test. f Nonlocal inhibition strengths for edge (open bars) and surface responses (filled bars) to black (black bars) and white square (white bars). Surface response to a white surface has the strongest inhibition. ***p < 0.001; n.s.: p = 0.9412; two-sided paired t-test. g Spatial extension for pooled excitation (red bars) and nonlocal inhibition (blue bars) for the black and white signal. ***p < 0.001, two-sided paired t-test. h Model performance with constraints for different spatial ranges of inhibition. Error bars show s.e.m. across sites. Source data are provided as a Source Data file.