Article Figures & Data
Figures
- Figure 1.
Example kinematics. A, Example hand paths of Monkey M reaching for cursor-on (top) and cursor-off trials (bottom). B–D, Example hand paths for goal jumps (B), cursor jumps (C), and mechanical loads (D). Solid and dashed lines are perturbations requiring corrections toward and away from the body, respectively. E, The average hand speed on cursor-on and cursor-off trials. F–H, The change in the lateral hand velocity for goal jumps (F), cursor jumps (G), and mechanical loads (H). Note, for the mechanical loads the change in lateral hand velocity starts at 0 ms due to the displacement caused by the loads. Arrows denote the onset of the kinematic correction using a difference function between the perturbations in opposite directions.
- Figure 2.
Movement times and endpoint distance from goal across monkeys. A, Movement times for Monkey M for cursor-on and cursor-off unperturbed reaches. Movement time was defined as the time between when the hand left the start target and when the hand first contacted the goal target. Trials have been pooled across all recording sessions. Arrows denote the median of the distributions. Distributions for cursor-on and cursor-off trials were not significantly different (two-sample t test: t(471) = 1.6; p = 0.12). B, Same as A for perturbation trials. C, Same as A except for the distance the reach endpoint was from the goal. Distributions for cursor-on and cursor-off trials were significantly different (t(471) = 3.6; p < 0.001). D, Same as C for perturbation trials. E–H, Same as A–D for Monkey A. E, Distributions for cursor-on and cursor-off trials were not significantly different (t(279) = 1.9; p = 0.06). G, Distributions for cursor-on and cursor-off trials were significantly different (t(279) = 4.0; p < 0.001). Note, Monkey M had longer movement times than Monkey A due in part to Monkey M completing a 10 cm reach and Monkey A completing an 8 cm reach.
- Figure 3.
Example neuron activities. A, Activities from four example neurons (first four rows) and muscle activity (bottom row) during reaches for cursor-on (black) and cursor-off trials (gray). Gray area demarcates when vision was removed. B–D, The change in activities (ΔActivity) for the same four example neurons and muscle in response to the goal jumps (B), cursor jumps (C), and mechanical loads (D). Solid and dashed lines are responses to perturbations requiring corrections toward and away from the body, respectively. Arrows denote when a significant change in activity was detected using the difference function between perturbations in the opposite direction.
- Figure 4.
M1 activity is largely unaffected by removing cursor feedback. A, For Monkey M, comparison of the mean activities during unperturbed reaches for cursor-on (abscissa) and cursor-off (ordinate) trials. Activity was averaged from 100 to 250 ms after the cursor feedback was removed. Each circle denotes one neuron. Dashed line reflects the line of best fit identified using total least squares regression (slope indicated in top left corner). B, Same as A except for the standard deviation across trials. C, D, Same as A and B except for Monkey A. E, F, Same as A and B except for EMG from Monkey M.
- Figure 5.
Proprioceptive feedback alters M1 activity earlier than visual feedback. A, The average activity across neurons for Monkey M. Arrows indicate when a significant increase from baseline was detected. Only neurons with significant activity for at least one perturbation type were included. B, The onset across individual neurons for the target jumps. Different colors of bars denote subsets of neurons. Dark green bars indicate neurons that responded to all three perturbation types (Fig. 6), light green bars denote neurons that responded to the target jump and only one other perturbation type (cursor jump or mechanical loads), and open bars denote neurons that responded to target jumps only. C, D, Similar to B for cursor jump and mechanical load onsets. E–H, Same as A–D for Monkey A. I–L, Same as A–D for muscle activity from Monkey M.
- Figure 6.
Each perturbation type influences overlapping neurons. A, Venn diagram showing the number of neurons observed (Obs) in each class for Monkey M. The diagram also shows the number of expected (Exp) neurons assuming an independent distribution. Chi reflects the classes contribution to the total χ2 value ([Obs-Exp]2/Exp). B, Venn diagrams classifying neurons using only two perturbation types for Monkey M. C, D, Same as A and B except for Monkey A.
- Figure 7.
M1 neurons have similar response ranges across perturbation types. A, Comparison of the response ranges between activities for the goal and cursor jumps. Black circles: neurons responsive to all three perturbation types. Gray circles: neurons responsive to at least one perturbation type. “r” is Pearson's correlation coefficient. Dashed lines reflect the line of best fit identified using total least squares regression (slope indicated in quadrant 2). B, Same as A except comparing mechanical loads and goal jumps. C, Same as A except comparing mechanical loads and cursor jumps. D–F, Same as A–C except for Monkey A. G–I, Same as A–C except for muscle activity from Monkey M.
- Figure 8.
Perturbation-related activity is comparable to activity during baseline reaching. A, Activities of the same four example neurons in Figure 2 during unperturbed reaches aligned to movement onset (5% max hand speed). Shaded area denotes the movement epoch (−50 to 250 ms). B, Scatter plot comparing the absolute magnitude of movement-related activity with the magnitude of the perturbation-related activity. C, Cumulative sums of the difference in the magnitudes of the movement-related and perturbation-related activities across cells.
- Figure 9.
Activity patterns overlap across perturbation types. A, Variance accounted for by the top goal–jump principal components for Monkey M. Data are presented as a cumulative sum showing how inclusion of each principal component increases the total variance captured for each perturbation type. Circles and bars denote the median and the 5th and 95th percentiles of the distributions. B, C, Same as A for cursor jumps and mechanical loads. D, Overlap index between perturbation types (clear bars) and the shuffle and within-perturbation distributions (filled bars). Bars denote the median and 5th and 95th percentiles of the distribution. E–H, Same as A–D except for Monkey A. I–L, Same as A–D except for EMG from Monkey M.
- Figure 10.
Time course of overlap index. A, Time series of the overlap index between goal and cursor jumps (black solid line) for Monkey M. Activity was binned every 20 ms. The time series was also repeated for the shuffle distribution (black dashed line) and the within-perturbation distributions for the goal-related (green line) and cursor-related (blue line) activities. B, Same as A except comparing mechanical loads with goal jumps. C, Same as A except comparing mechanical loads with cursor jumps. D–F, Same as A–C except for Monkey A. G–I, Same as A–C for EMG signals. Prior to overlap calculation, EMG signals were filtered with a low-pass third-order Butterworth filter (cutoff 50 Hz). Note, the substantial overlap before perturbation onset is in part due to the small subspace spanned by EMG signals.
- Figure 11.
Overlap across perturbation types with increased perturbation directions. A, Monkey M's lateral reaches following goal jumps (left), cursor jumps (middle), and mechanical loads (right). Same as Figure 1B–D. B, Same as A except now for Monkey's M anterior reaches. C, Response ranges comparing perturbation types for the anterior reaches. Data presented the same as in Figure 5. “n” denotes the number of recorded neurons. D, Overlap index presented the same as Figure 6. E, F, Same as C and D for Monkey A.
Extended Data
Table 1.
Onsets for M1 neurons identified as responsive to target-only, cursor-only and mechanical-only perturbations. Neurons were pooled across both monkeys. Obs: observed, Exp: expected. Download Table 1, DOCX file.
Figure 1.
M1 neurons have similar response ranges across perturbation types during the onset of evoked activity. Data are presented the same as Figure 7 except response ranges were calculated for the visual perturbations from 75-125ms after perturbation onset and for the mechanical perturbations from 50-100ms after the perturbation onsets. Download Figure 1, TIF file.
In this issue
