Fig. 5: Near-orthogonal relationship between decision- and stimulus-related subspace.
From: Domain-specific schema reuse supports flexible learning to learn in the primate brain
a The angle between the stimulus- and the decision-related subspace was compared with angles obtained after shuffling the decision and stimulus labels. Data are shown as box-and-whisker plots. The line within each box indicates the median; box edges represent the first (Q1) and third (Q3) quartiles; whiskers show the full data range. Colored scatter dots represent individual sessions, with each color corresponding to one monkey. (Original group: n = 3 per monkey; Shuffled group: n = 30 per monkey; Monkey ZZ: P = 0.04, g = 1.27; Monkey AB: P = 0.005, g = 1.77; Monkey XW: P = 0.002, g = 2.02). b Distribution of data for individual trials of monkey AB on the stimulus-related and decision plane. c Illustration of the angle between the decision- and stimulus-related subspaces. Blue and orange areas represent decision- and stimulus-related subspace distributions, respectively. Red and green dashed lines mark example angles of 81.4° and 25°, with a gray right angle indicating near-orthogonality. d Neural trajectories in decision- and stimulus-related subspaces at different angles. Left: Neural trajectories and stimulus projections when the subspace angle is 81.4°. Right: Neural trajectories for a selected case at 25°. Lines represent trial-averaged trajectories, and shaded areas represent the variability across trials for each stimulus. *, P < 0.05, **, P < 0.01, ***, P < 0.001, Watson–Williams test.
