Article Figures & Data
Figures
- Figure 1.
A, Four examples of closure. The yellow dots and ellipses indicate the center and extent of the CRFs, respectively. The closure values were the percentages of the radial lines extending from the CRF center that crossed the contour. The gray dots are the points on the contour. B, Four examples of curvature. The gray dots are the points on the contour, and the gray regions indicate the figure regions. The small dots and ellipses in black indicate the center and extent of the CRFs. The blue squares indicate the windows for the computation. The pink lines indicate the circles fitted to the contours within the CRF and window. Refer to the main text and Extended Data Figure 1-2 for details. C, Four examples of symmetry. The gray dots indicate the sampled points on the contour. The red lines are the mid-line of the stimulus used for the computation (vertical in these cases). D, Four examples of orientation. Left panels, The red dots and ellipses indicate the center and extent of the CRFs. The gray dots are the points on the contour. Right panels, The histograms of the local orientations within the CRF. The red lines on the histograms indicate the mean local orientation, which are the orientation values for the combinations of the neuron and stimulus. Example tuning maps for closure (E), curvature (F), symmetry (G), and orientation (H). The asterisks indicate the significance of tuning. The bottom left schematics illustrate the degree of the feature. I, Example time courses of 10 neurons that showed significant tuning. The original stimuli and their correlation between the closure and curvature values are shown in Extended Data Figures 1-1 and 1-3, respectively.
- Figure 2.
A, Example tuning curves for two neurons computed from the responses to all stimuli including natural and silhouette stimuli. Arrows represent the peaks of the individual tuning curves. B, The probability distributions of the four contour features of the stimulus set. C, The distributions of the optimal feature values of the neurons. The arrows correspond to the peaks in the individual tuning maps shown in A. D, The normalized distributions of the optimal feature values; C was normalized by B.
- Figure 3.
A, The histogram of the number of significant features of individual neurons. The detailed numbers are shown in Extended Data Figure 3-1. B, The distribution of the normalized tuning strengths of the optimal closure and curvature, with the enlargement in the top right inset. The blue, red, and green dots indicate the neurons with significant tuning for closure, curvature, and both of them, respectively. C, The distribution of the cross-tuning between closure and curvature (the combination of optimal closure and curvature) of individual neurons that exhibited significance to closure and/or curvature. The colors represent the number of neurons. D, The distribution of closure and curvature values across the stimuli. E, The normalized probability distribution of the cross-tuning; C was normalized by D. The colors represent the probability. F–H, The same plots as C, D, and E, respectively, for the neurons with significant tuning for both closure and curvature.
- Figure 4.
A, Top, The histograms for the MIclos. The orange bars indicate the distributions of neurons with significant tuning to closure and blue bars all neurons. The orange and blue vertical lines indicate the mean MI values for the significant and all neurons, respectively. Bottom panels, Example tuning maps of neurons with small, medium, and large MI values (from left to right), with the conventions same as Figure 1E. B–D, The histograms and example tuning maps for MIcurv, MIsymm, and MIorien, respectively. E, The combinations of MIclos, MIcurv, MIsymm, and MIorien for individual neurons in the order of the sum of the four MIs. The colors indicate the types of MI. F, The ratios of the greatest MI with respect to the sum of the four MIs. The conventions are the same as E. The horizontal dotted line indicates the even distribution (0.25). G, The ratios of the least MI with respect to the sum of the four MIs. The conventions are the same as F. The analyses including the neurons with a small number of spikes are shown in Extended Data Figure 4-1.
- Figure 5.
A, Top, The two-dimensional MDS map generated from both natural and silhouette stimuli. Refer to Extended Data Figure 5-1 for the stress. The dots represent individual neurons. The neurons are divided horizontally into five groups (as shown in different colors) so as to include the equal number of neurons. Stimuli that evoked the strong responses to example neurons are shown in Extended Data Figure 5-2. Middle panel, The ratios of neurons with significant tuning with moving average (refer to Materials and Methods for details). The blue, orange, yellow, and purple lines show the ratios of significance to closure, curvature, symmetry, and orientation, respectively. The abscissa corresponds to that of the MDS map on the top. The shades represent standard deviation. Bottom panel, The mean number of significant contour features along with the primary axis of the MDS map, together with the normalized mean optimal closure and curvature in blue and orange lines, respectively. B, The MDS map generated from the responses to silhouette stimuli (top panel), the ratios of the significantly tuned neurons (middle panel), and the mean number of significant contour features together with the mean optimal closure and curvature (bottom panel). The conventions are the same as A. C, Left panel, The generated MDS map identical to A but with vertical grouping. Right panel, NS preference (refer to Materials and Methods). The ordinate corresponds to that of the MDS map. The blue line indicates the moving mean of the NS preference and the shades represent standard deviation.
Extended Data
Extended Data Figure 1-1
The natural and silhouette stimuli (the left and right columns, respectively). Their mirror images were also presented in the experiments (refer to the Methods for details) but not shown here. Download Extended Data Figure 1-1, TIF file.
Extended Data Figure 1-2
A schematic diagram illustrating the computation of curvature and the generation of the response histogram used for the estimation of tuning curve. (A) An illustration showing the combinations of a stimulus and neurons for the computation of curvature. The value of curvature within a single stimulus varies depending on neurons since the CRF extent depends on neurons. The panels share the presentation of the same contour (a black/red curve) but for different neurons with the distinct extents of the CRF (gray ellipses). For example, the leftmost panel shows the combination of neuron 1 and stimulus 1, with enlargement. Red parts of the contours indicate those within the CRFs wherein local curvatures are calculated. (B) An illustration for the computation of curvature for the combination of stimulus 1 and neuron 1. A set of windows with various scales (1, 3/4, 2/4, 1/4, 1/10) and locations (blue squares; the common fractions next to the windows indicate the scale with respect to the stimulus) was given for the computation of local curvatures. A single local curvature was calculated for each single window. A circle (a pink curve) was fitted to the contour within the window and CRF (a blue curve). Only a central part of the stimulus was illustrated in the lower panels surrounded by a dotted square which corresponded to the dotted square in the top-left panel. Depending on the location and scale, a different part of contour was fitted by a circle, and thus a different local curvature was obtained. (C) The left panel shows an example histogram of local curvatures for the combination of neuron 1 and stimulus 1, which indicates the distribution of local curvatures in this stimulus. The response histogram for this single stimulus was given by multiplying the spike response of neuron 1 to stimulus 1 (in this example, the response was 2). In short, this histogram shows how the response was weighted by the local curvatures. By adding these response histograms across all stimuli, the response histogram for this neuron was obtained. Download Extended Data Figure 1-2, TIF file.
Extended Data Figure 1-3
The correlation between the closure and curvature values across stimuli (R=0.0017, p=0.50). Only randomly-chosen 100 data are shown here for a presentation purpose. The insets are example stimuli. The “F” indicates the figure region. Download Extended Data Figure 1-3, TIF file.
Extended Data Figure 3-1
The Venn diagram indicating the numbers of neurons with the significant tuning for the contour features. Download Extended Data Figure 3-1, TIF file.
Extended Data Figure 4-1
(A) The blue line shows the correlation between the MIclos and the number of neural spikes. The red lines represent 0.2, 0, and -0.2. (B), (C), and (D) show the correlations for MIcurv, MIsymm, and MIorien , respectively, with the conventions same as (A). (E), (F), (G), and (H) are the histograms of MIclos, MIcurv, MIsymm, and MIorien, respectively, without excluding neurons with a small number of spikes. The orange bars show those with signicance, and blue show those with and without significance. (I) The combinations of MIclos, MIcurv, MIsymm, and MIorien for individual neurons including those with a small number of spikes in the order of the sum of the four MIs. The colors indicate the types of MI. (J) The ratios of the greatest MI with respect to the sum of the four MIs. The conventions are the same as (I). The horizontal dotted line indicates the even distribution (0.25). (K) The ratios of the least MI with respect to the sum of the four MIs. The conventions are the same as (J). Download Extended Data Figure 4-1, TIF file.
Extended Data Figure 5-1
(A) and (B) are the stress plots for the MDS in Figure 5A and 5B, respectively. Download Extended Data Figure 5-1, TIF file.
Extended Data Figure 5-2
Small image patches are the stimuli that evoked the five strongest responses of the example neurons, in correspondence with the MDS map (Figure 5A). The neurons on the left and right sides tended to evoke strong responses to relatively simple and complex stimuli, respectively. Download Extended Data Figure 5-2, TIF file.
In this issue
