Fig. 3: Neural correlates of behavior vary across tasks in PMd, but stabilize in M1 and even more in S1.

a We calculated single neuron firing rates during each gait cycle, defined as the epoch between two consecutive foot strikes of the right hindlimb contralateral to neural recordings. We then identified the right hindlimb stance and swing phases for each gait cycle, and then time-warped these periods to 60% and 40% of the gait cycle. We averaged these warped trials across all gait cycles to obtain peri-gait neuronal firing rates for each task. b The colorplots show peri-gait firing rates of isolated neurons aligned at stance and swing onsets and sorted by their preferred gait phase on each task separately. The neurons recoded in different sessions are grouped together for visualization purposes. All statistics shown later are calculated by treating different sessions as separate data sets with no assumptions about whether the neurons recorded in different sessions are the same or different. c Same data as (b) with the neurons sorted by their preferred gait-phase in the corridor task. Note the substantial changes in neural activity, reflected by changes in gait phase tuning for each neuron. d Left panels show peri-gait firing rates of one representative neuron from PMd, M1, and S1. Similarity across tasks indicates that the recorded signals came from the same neuron. For every neuron, we computed the Pearson’s linear correlation coefficient (R) between the activity on any two gait cycles belonging to different tasks. We then averaged these values across all gait cycle pairs in a particular task combination (10 combinations), all neurons from a cortical region, and all task combinations. The bars show the population cross-task correlation as the mean of these cross-task region-specific correlation values across all sessions and both monkeys (n = 14; PMd: 0.42 ± 0.03; M1: 0.67 ± 0.02; S1: 0.71 ± 0.01). Each dot shows the region-specific mean correlation value for one session of one monkey (dots on the left column, Mk-Ek; dots on the right column, Mk-Nt). Black line shows the chance level with the grey tube showing the values not significantly different from chance at p ≥ 0.05 (chance level: PMd: 0.0002 ± 0.009; M1: 0.0000 ± 0.0013; S1: 0.0000 ± 0.001; measured value vs. chance: PMd: p = 0.0005; M1: p = 0.0005; S1: p = 0.0005). Blue line shows the estimated noise contribution with light blue tube showing the values not significantly different from noise contribution at p ≥ 0.05 (PMd: 0.768 ± 0.009; M1: 0.885 ± 0.005; S1: 0.896 ± 0.005; measured value vs. noise contribution: PMd: p = 0.0005; M1: p = 0.0005; S1: p = 0.0005). Population cross-task correlation was lower in PMd than in M1 and S1 (PMd vs. M1: p = 0.0005; PMd vs. S1: p = 0.0005; M1 vs. S1: p = 0.13). e Left panel shows the peri-gait firing rates for one representative neuron with its preferred gait phase (PGP) identified as the phase of maximum activity. We identified each neuron’s PGP for each task and computed the circular standard deviation of PGP across tasks. The bars show the population PGP standard deviation as the mean PGP circular standard deviation across all neurons of a cortical region, all sessions and both monkeys (n = 14; PMd: 42.7 ± 1.7; M1: 37.1 ± 1.2; S1: 23.9 ± 1.2). Each dot shows the mean value for each session and each monkey (dots on the left column, Mk-Ek; dots on the right column, Mk-Nt). Black line shows the chance level with the grey tube showing the values not significantly different from chance at p ≥ 0.05 (chance level: 60.33 ± 10.85; measured value vs. chance: PMd: p = 0.0005; M1: p = 0.0005; S1: 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: PMd: 18.4 ± 0.9; M1: 14.5 ± 0.9; S1: 10.0 ± 0.4; measured value vs. noise contribution: PMd: p = 0.0005; M1: p = 0.0005; S1: p = 0.0005). Population PGP circular standard deviation trended smaller in M1 compared to PMd, and was smaller in S1 compared to M1 (PMd vs. M1: p = 0.058; PMd vs. S1: p = 0.0005; M1 vs. S1: p = 0.0005). Error bars: s.e.m.; n.s. p ≥ 0.05; *** p < 0.001; one-sided Monte Carlo Permutation test. Source data are provided as a Source Data file.