Fig. 6: Task-independent subspaces are preferentially aligned to communication subspaces across cortical regions, except those targeting PMd.
From: Regional specialization of movement encoding across the primate sensorimotor cortex
a Step 1: We used a reduced rank regression algorithm to identify the source communication subspaces. Step 2: We computed the principal angles between the source communication subspaces and source task-independent subspaces (CS-TI angles), as well as the source communication subspaces and source task-dependent subspaces (CS-TD angles). We then computed the angle difference (∆ angles), defined as the difference between the mean CS-TD angles and the mean CS-TI angles. b The plots show the distribution of all the ∆ angles (n = 35: 5 tasks x 7 sessions) for every combination of source and target cortical regions and for each monkey. Positive ∆ angles values show preferential alignment with the task-independent subspaces. c Barplots show the mean of ∆ angles across all tasks and sessions for every combination of source and target cortical regions and for each monkey (n = 7; Mk-Ek: PMd to M1: 6.80 ± 1.49; PMd to S1: 5.52 ± 1.04; M1 to PMd: -4.82 ± 1.53; M1 to S1: 2.36 ± 0.69; S1 to PMd: −3.55 ± 1.54; S1 to M1: 3.28 ± 0.74; n = 7; Mk-Nt: PMd to M1: 7.28 ± 1.20; PMd to S1: 8.17 ± 1.14; M1 to PMd: −2.75 ± 1.10; M1 to S1: 5.97 ± 0.60; S1 to PMd: −1.62 ± 0.72; S1 to M1: 6.43 ± 2.35). Dots show the mean of ∆ angles across all tasks for every session. Black line shows the chance level with the grey tube showing the values not significantly different from chance at p ≥ 0. 05 (chance level: Mk-Ek: PMd to M1: 0.05 ± 0.67; PMd to S1: 0.06 ± 0.69; M1 to PMd: 0.03 ± 0.65; M1 to S1: 0.00 ± 0.65; S1 to PMd: 0.04 ± 0.56; S1 to M1: 0.06 ± 0.56; Mk-Nt: PMd to M1: 0.06 ± 0.57; PMd to S1: 0.08 ± 0.57; M1 to PMd: 0.02 ± 0.58; M1 to S1: 0.06 ± 0.56; S1 to PMd: −0.02 ± 0.74; S1 to M1: −0.05 ± 0.75; measured value vs. chance level: Mk-Ek: PMd to M1: p = 0.0005; PMd to S1: p = 0.0005; M1 to PMd: p = 0.0005; M1 to S1: p = 0.0009995; S1 to PMd: p = 0.0005; S1 to M1: p = 0.0005; Mk-Nt: PMd to M1: p = 0.0005; PMd to S1: p = 0.0005; M1 to PMd: p = 0.0005; M1 to S1: p = 0.0005; S1 to PMd: p = 0.02; S1 to M1: p = 0.0005). Blue line shows the estimated noise contribution with the light blue tube showing the values not significantly different from noise contribution at p ≥ 0.05 (noise contribution: Mk-Ek: PMd to M1: −1.56 ± 0.60; PMd to S1: −1.59 ± 0.57; M1 to PMd: −11.07 ± 0.50; M1 to S1: −11.33 ± 0.51; S1 to PMd: −8.63 ± 0.40; S1 to M1: −8.90 ± 0.40; Mk-Nt: PMd to M1: −0.56 ± 0.47; PMd to S1: −0.47 ± 0.44; M1 to PMd: −8.39 ± 0.50; M1 to S1: −6.70 ± 0.48; S1 to PMd: −9.14 ± 0.69; S1 to M1: −9.18 ± 0.68; measured value vs. noise contribution: Mk-Ek: PMd to M1: p = 0.0005; PMd to S1: p = 0.0005; M1 to PMd: p = 0.0005; M1 to S1: p = 0.0005; S1 to PMd: p = 0.0005; S1 to M1: p = 0.0005; Mk-Nt: PMd to M1: p = 0.0005; PMd to S1: p = 0.0005; M1 to PMd: p = 0.0005; M1 to S1: p = 0.0005; S1 to PMd: p = 0.0005; S1 to M1: p = 0.0005). Error bars: s.e.m.; * p < 0.05; *** p < 0.001; one-sided Monte Carlo permutation test. Source data are provided as a Source Data file.
