Same mutual information analysis as the one shown in Figure 3, but considering separately the neuronal populations recorded in cortical layers II-IV (left plots) and V-VI (right plots). Colors, symbols, significance levels and statistical tests as in Figure 3. (A) To check whether the drop of stimulus information (full bars) was similarly sharp in superficial and deep layers, a two-way ANOVA, with visual area and layer as factors, was carried out. The test yielded a significant main effect for area (p<0.001, F3,687 = 15.87) and a significant interaction between area and layer (p<0.001, F3,687 = 5.61), but not a main effect for layer alone (p>0.9, F1,687 = 0.01). This indicates that the information loss was sharper in deep layers. As for the case of the entire populations (see Figure 3B), the drop of information was mainly due to a reduction of the information about stimulus luminosity (white bars), while the information about higher-order visual attributes (colored bars) remained more stable across the areas. This was especially noticeable for deep layers, where the information about luminosity in LM and LI became almost half as large as in V1, further dropping in LL to about one third of what observed in V1. (B) The fraction of luminance-independent stimulus information (fhigh) that neurons carried in each area increased more gradually and was, overall, larger in deep than in superficial layers. A two-way ANOVA, with visual area and layer as factors, confirmed this observation, yielding a significant main effect for both area (p<0.001, F3,687 = 63.58) and layer (p<0.001, F1,687 = 14.7) and also a significant interaction (p<0.01, F3,687 = 4.06). In terms of pairwise comparisons, this resulted in a very large and significant increase of fhigh in LL, as compared to all other areas, in both superficial and deep layers, and in a significant difference also between LI and V1/LM, but in deep layers only.