Fig. 6: Orientation (ori) bandwidth as a function of preferred SF (fo). | Nature Communications

Fig. 6: Orientation (ori) bandwidth as a function of preferred SF (fo).

From: Uniform spatial pooling explains topographic organization and deviation from receptive-field scale invariance in primate V1

Fig. 6

a The scatter plot compares fo to ori bandwidth in three ROIs, each indicated by a different symbol. See Table 2 for joint statistics. The five data points at the extreme left of the x-axis were low pass in SF. The blue line is the ori bandwidth based on a scale invariant aspect ratio of 2.0. The green line is the model of pooled scale invariance, which did not require any additional parameters. Rather, it was derived from the fit to SF bandwidth (Fig. 5a, green). To the left of the y-axis are ori tuning curves (black dots) and Gaussian fits (gray) of four example cells in one ROI outlined and enumerated in e. b Each of the three images represents the average response from a different subpopulation, as a function of SF (x-axis) and ori (y-axis). Blue-to-red represents normalized response from min-to-max, for each panel. The image on the right is the mean over cells with fo > 2 cyc/°, the middle image is over cells with 1 < fo < 2 cyc/°, and the left image is over fo < 1 cyc/°. Curves on the left are the mean ori tuning curves computed from the three images on the right; they are averaged over the SF dimension. Blue, red, and yellow are for the images from left to right, respectively. c Simulation of the pooling model in the 2D spectral domain, at fo = 1.5 and 6 cyc/°. At bottom is an angled perspective of the 2D Fourier plane, where the color represents orientation and fo increases linearly away from the center. Each location of the 2D Fourier plane corresponds to a different 2D scale invariant Gaussian. Pooling closer to the origin (@ 1.5 cyc/°) yields broader ori tuning than pooling further away from the origin (@6 cyc/°), assuming an invariant pooling function (σh(f) = 0.85 cyc/°). The simulation on top shows a random sample of weighted and shifted ori tuning curves (blue) from the two indicated pooling locations, along with their superposition (green). d Modeled maps of ori bandwidth. Bottom and top panels are generated by plugging fo into the scale invariance and pooled scale invariance model, respectively. e Measured maps of ori bandwidth. f Pearson correlation coefficient between neural pairs, within 75 μm cortical distance bins, using all three ROIs in the study. The symbol at each distance bin indicates the correlation coefficient’s significance: Closed dot is p > 0.01, asterisk is p < 0.01, asterisk and open circle is p < 0.001.

Back to article page