Fig. 4: Changed fractions of stimulus- and task-activated neurons after category learning.
From: Mouse visual cortex areas represent perceptual and semantic features of learned visual categories
a, Fraction of responsive neurons (corrected for variable trial numbers by subsampling). The y axis shows chronic recordings, and the x axis shows imaging time points. TC, out-of-task time points (tuning curves). Task, in-task time points. L, separates baseline from after category learning. The second in-task baseline was only acquired in a subset of mice. b, Inertia of k-means identified clusters of chronic recordings, as a function of the number of initialized clusters (k). Black, real data; gray, shuffled data. The inset shows Δinertia of real and shuffled data. The arrow indicates maximum Δinertia (at two clusters). c, Fraction of responsive neurons for each cluster of chronic recordings (gray represents cluster 1, and blue represents cluster 2). Bars show the mean (±s.e.m.), and light-gray lines show individual chronic recordings (n = 39 chronic recordings from ten mice). d, Map of mouse visual areas (based on ref. 37) showing the fraction of cluster 1 and cluster 2 chronic recordings per area. e, Number of chronic recordings per area, color coded for cluster identity. D, dorsal stream-associated area; V, ventral stream-associated area. Areas V1 and PM are equally associated with both streams and are therefore unlabeled (n = 39 chronic recordings from ten mice). f, The fraction of cluster 1 and cluster 2 recordings in dorsal stream-associated areas AL, RL and AM, and ventral stream-associated areas LM, LI, P and POR (chi-squared test, χ2(1) = 8.58, P = 0.0034; ndorsal = 12, nventral = 15 chronic recordings from ten mice). g, Schematic of linear model predicting the fraction of responsive neurons per chronic recording (Methods). Individual components; base, overall non-changing fraction of responsive neurons; decay, exponential session-dependent reduction; task, increase during in-task time points; learning, increase after category learning. h, Weight (mean, across chronic recordings) of each component per visual cortical area (B, base; D, decay; T, task; L, learning). i, Mean (±s.e.m.) weight associated with each model component, separately for chronic recordings (gray dots) from dorsal and ventral stream-associated areas (two-sided Mann–Whitney U test; base, U = 30, P = 0.0073; decay, U = 89, P = 1.96; task, U = 79.0, P = 1.22; learning, U = 36, P = 0.013; ndorsal = 12, nventral = 15 chronic recordings from ten mice; P values were calculated using Bonferroni correction for four comparisons). NS, P > 0.05, *P < 0.05, **P < 0.01.
