Fig. 4: Intact whisker input and barrel cortex are necessary for normal task performance and learning.

a Population performance for expert animals before and after bilateral whisker-pluck. Performance calculated from last session before and first session after the whisker pluck, each dot represents one session, dashed line indicates expert-level threshold (d’ = 1.65), n = 6 mice. b Same as in a but with unilateral whisker pluck and for contrasts of 6 and 20 mm (n = 4 mice for each contrast). c Same as in a but with barrel cortex ablations (red) or sham operations (gray). d Performance over trials in barrel cortex ablated mice (red) and sham operated animals (gray). Expert-level performance (d’ = 1.65, gray horizontal dashed line) was crossed after 478 vs. 2023 trials for sham (gray vertical dashed line) and ablated mice (red vertical dashed line), each n = 4 mice. Inset: example reconstruction of barrel cortex ablation. e Population performance for expert animals in d, calculated from the session with the maximum d prime for each mouse (n = 4 mice each). f Proportion of aperture-selective BC units across learning progression (solid line, mean ± SEM). Aperture selectivity was assessed using a two-sided Wilcoxon signed-rank test, comparing unit firing rates between the two apertures. The dotted light blue line represents the average d-prime (± SEM), n = 6 mice. g Population performance for expert animals (n = 4 mice) with alternating application of lidocaine or neutral ointment (control) to both whisker pads. Performance was calculated from 7 lidocaine and 7 control sessions across all mice for each condition. Same conventions as in a. h Time-resolved decoding accuracy and standard error of mean (bands) of aperture width from spike data of different brain areas. Spike data was analyzed from expert sessions (800 ms before to 400 ms after aperture touch, n = 6 mice). i Receiver operating characteristic curves from a generalized additive model classifier based on whisker angles after whisker-touch onset. Comparison of control (black), lidocaine sessions (orange), and randomly shuffled labels (light gray). Accuracies for control and lidocaine sessions were not significantly different from each other, but both differed significantly from the shuffled accuracies (n = 3 mice). j Decoding accuracy from spike data for lidocaine sessions (orange frames) and control sessions (black frames). The mean decoding accuracy was calculated over a time window from trigger onset to 400 ms post-trigger onset (see Methods). Number of units (from n = 3 mice) in lidocaine sessions: 128 (BC), 100 (VPM), 52 (POm), 17 (ZIv); and control sessions: 150 (BC), 101 (VPM), 64 (POm), 17 (ZIv). ns: p > 0.05, *p ≤ 0.05, **p ≤ 0.01 and ***p ≤ 0.001; a, b two-tailed paired t-test, c two-way anova, e independent samples t-test, g repeated measures anova, j two-tailed Wilcoxon rank sum test. For exact statistics and box plot definitions see Supplementary Table 1.