The percentage of unimodal cells decreased with the distance from the border of the respective
primary area (solid lines in Figure 5C, left). Conversely, bimodal cells were uniformly distributed, with a slight increase in the middle of the field of view. To test for a gradient in the density of unimodal selleck screening library cells along the V1-S1 axis, we performed a linear regression on the cell-density values (dashed lines in Figure 5C, left) For unimodal cells, the slopes were significantly different from zero (see Figure 5C, right; slopes: −0.066 for T cells, p = 0.017; 0.078 for V cells, p = 0.005; permutation test for the slope; see Supplemental Experimental Procedures). The distribution of bimodal cells did not show a spatial gradient
(−0.012 for M-driven cells, p = 0.68; permutation test for the slope). Moreover, we failed to find a similar gradient for a modality dominance index—which expresses the relative strengths of the two modalities—computed on responses of bimodal cells (Figure S4B). This indicated that bimodal cells near one primary cortex were not functionally dominated by the corresponding modality. We then wondered whether the three types of responsive neurons showed some kind of spatial clustering on a microscale level. Since the gradient of unimodal neurons could be see more a confounding factor, we restricted our analysis to the middle stripe of the imaged area (i.e., a portion of RL oriented orthogonal to the V1-S1 axis and equidistant from both S1 and V1—see also additional controls in Table S2). Within this stripe the mean position and density of cell
somata along the V1-S1 axis were statistically indistinguishable for V, T, and M cells, indicating a homogenous distribution of unimodal neurons within the central cortical stripe (Figures S4C and S4D). We performed a nearest-neighbor analysis in the center of area RL for V, T, and M cells separately on single optical planes (Komiyama et al., 2010; Figure 5D). As the three cell types had a similar density, we took 0.33 as a chance probability for the nearest PAK6 neighbor analysis (Figure 5E). For each cell type, we first computed the probability of having a nearest neighbor of a certain type. For unimodal cells, the probability that the nearest neighbor was another unimodal neuron of the same modality was above chance (Figure 5E; for T cells: 52.5% of T neighbors, p < 0.001; for V cells: 55.7% of V neighbors, p < 0.001) and the probability that the nearest neighbor was a unimodal neuron but driven by the other modality was below chance (for T cells: 17.1% of V neighbors, p < 0.01; for V cells: 20.0% of T neighbors, p < 0.05). For unimodal cells, the probability that the nearest neighbor was bimodal did not differ from chance. Conversely, for bimodal cells, the nearest neighbor could either be a T, V, or M cell, with a trend toward M cells (29.4% of T neighbors, p = 0.58, 30.9% of V cells, p = 0.27, 39.