Fig. 6: Category selectivity emerges in choice-selective POR neurons.
From: Mouse visual cortex areas represent perceptual and semantic features of learned visual categories
a, Schematic with two examples, demonstrating that feature CTI is calculated from orientation/spatial frequency-specific weight kernels, and semantic CTI from category-specific weight kernels (obtained using the full GLM; Extended Data Fig. 8). Colored curves represent the tuning curve of a neuron (red, right-category stimuli; blue, left-category stimuli; in the grid, the x axis shows the orientation, and the y axis shows spatial frequency). b, Mean semantic CTI for all neurons, pooled across mice and areas, before and after category learning (two-sided WMPSR test, W = 86171, P = 1.83 × 10−4; n = 645 neurons from ten mice). Gray dots indicate individual neurons (Extended Data Fig. 9a). c, The difference in semantic CTI before (‘baseline 3’) and after (‘learned 1’) category learning, overlaid on a map of mouse visual areas (based on ref. 37; two-sided WMPSR test; V1, W = 4,088, P = 0.036; n = 149 neurons from seven mice; POR, W = 204, P = 0.009; n = 43 neurons from six mice; P values were calculated using Bonferroni correction for eight comparisons). Area P had no ‘stable’ visually modulated neurons, and area A was not imaged. d,e, As in b and c, for feature CTI (Extended Data Fig. 9b; pooled data (d), two-sided WMPSR test, W = 93043, P = 0.019; n = 645 neurons from ten mice). f, Mean ΔCTI, the difference between semantic CTI and feature CTI, before (triangle) and after (square) category learning, per visual area. Gray dots indicate individual neurons (Extended Data Fig. 9c; area POR, two-sided WMPSR test, W = 155, P = 9.85 × 10−4; n = 43 neurons from six mice; P values were calculated using Bonferroni correction for eight comparisons). g, Experimental timeline showing the stimulus discrimination session in which choice selectivity was determined (using the GLM), relative to sessions in which ΔCTI was calculated (‘baseline 3’ (bs3) and ‘learned 1’ (lrn1)). h, Correlation between choice selectivity at the time point of stimulus discrimination and ΔCTI after category learning (at ‘learned 1’). The black line represents the linear fit. Gray dots indicate individual neurons from all areas. Pearson correlation coefficient, r = 0.12, two-sided P = 0.0034; n = 620 neurons from ten mice. i, Mean (±s.e.m.) choice selectivity in POR neurons during initial stimulus discrimination, for neurons that increased (red circle) and decreased (blue circle) ΔCTI after category learning, compared to baseline (gray indicates individual neurons; two-sided, Mann–Whitney U test, U = 115, P = 0.047; n<bs = 12, n>bs = 29 neurons from five mice). *P < 0.05, **P < 0.01. bs, baseline.
