Neural population dynamics during reaching

Abstract

Most theories of motor cortex have assumed that neural activity represents movement parameters. This view derives from what is known about primary visual cortex, where neural activity represents patterns of light. Yet it is unclear how well the analogy between motor and visual cortex holds. Single-neuron responses in motor cortex are complex, and there is marked disagreement regarding which movement parameters are represented. A better analogy might be with other motor systems, where a common principle is rhythmic neural activity. Here we find that motor cortex responses during reaching contain a brief but strong oscillatory component, something quite unexpected for a non-periodic behaviour. Oscillation amplitude and phase followed naturally from the preparatory state, suggesting a mechanistic role for preparatory neural activity. These results demonstrate an unexpected yet surprisingly simple structure in the population response. This underlying structure explains many of the confusing features of individual neural responses.

Figure 1

Oscillation of neural firing rates during three movement types. a. Response of one of 164 neurons (simultaneously recorded using voltage-sensitive dye) in the isolated leech central nervous system during a swimming motor pattern. Responses (not averaged across repetitions) were filtered with a 100 ms Gaussian kernel. b. Multi-unit response from one of 96 electrodes implanted in the arm representation of caudal premotor cortex. Data from 32 such channels was wirelessly transmitted during walking. Responses (not averaged across repetitions) were filtered with a 100 ms Gaussian kernel. c. Response of one of 118 neurons recorded from motor cortex of a reaching monkey (N) using single-electrode techniques. Firing rates were smoothed with a 24 ms Gaussian and averaged across 9 repetitions of the illustrated leftwards reach (flanking traces show SEM). d. Projection of the leech population response into the 2-dimensional jPCA space. The two dimensions are plotted versus each other (top) and versus time (bottom). Units are arbitrary but fixed between axes. e. Similar projection for the walking monkey. f. Similar projection for the reaching monkey.

Figure 2

Firing rate versus time for 10 example neurons, highlighting the multiphasic response patterns. Each trace plots mean across-trial firing rate for one condition, colored based on the firing rate at the end of the preparatory period. Data were averaged separately locked to target onset, the go cue, and movement onset. To aid viewing, traces are interpolated across the gaps between epochs. Vertical scale bars indicate 20 spikes/s. Insets plot hand trajectories, which are different for each dataset.

Figure 3

Projections of the neural population response. a. Projection for monkey B (74 neurons; 28 straight-reach conditions). Each trace (one condition) plots the first 200 ms of movement-related activity away from the preparatory state (circles). Traces are colored based on the preparatory-state projection onto jPC₁. b. Projection for monkey A (64 neurons; 28 straight-reach conditions). c. Monkey J3 (55 neurons; 27 straight- and curved-reach conditions). d. Monkey N (118 neurons; 27 straight- and curved-reach conditions). e. Monkey J-array (146 isolations; 108 straight- and curved-reach conditions). f. Monkey N-array (218 isolations; 108 straight- and curved-reach conditions).

Figure 4

Projections of simulated neural and muscle populations. a. Simulated ‘velocity model’ unit, based on hand velocities of monkey A. The preferred direction points up and right. Presentation and scale bars as in figure 2. b. Simulated unit from the ‘complex kinematic model’. c. EMG from the deltoid (monkey J3), colored by initial response. EMG was rectified, smoothed, and averaged across trials. d. Projection of the ‘velocity model’ population response (64 simulated neurons) for monkey A. Identical presentation and analysis as figure 3. e. Projection of the ‘complex-kinematic model’ population response for monkey A. f. Projection of the recorded muscle population response for monkey A. h. Same as d but for monkey J-array. i. Same as e but for monkey J-array. j. Same as f but for monkey J3.

Figure 5

Illustration of how a simple model generates fits to EMG using a pair of brief rotations. a. The higher frequency rotation (2.8 Hz) for conditions 9 and 25, plotting the first 200 ms of evolution away from the preparatory state (filled circles). The preparatory state determines rotation amplitude and phase. One state dimension (‘leading’, on the x-axis) is always 90° ahead of the other (‘lagging’, on the y-axis). b. Projections onto the leading and lagging dimensions (light and dark blue) versus time (condition 9). The fit to the EMG is the sum of lagging components from the 2.8 Hz rotation (shown) and a lower frequency rotation (0.3 Hz, not shown). c. Similar to b, but for condition 25. The 2.8 Hz rotation is ~180° out of phase with that in b. d. Simulated response of a unit from the generator model (methods). Presentation as in figure 2. e. A second simulated unit. f. jPCA projection of the simulated population; compare with the neural data in figure 3c. g. Hand velocities for the 27 conditions in d, e, f. Red traces show conditions 9 and 25.

Figure 6

Consistency of rotational dynamics for real and simulated data. a. Histograms of the angle between the neural state, x, and its derivative, ẋ, for real and simulated data. The angle was measured after projecting the data onto the first jPCA plane as illustrated schematically (inset). Pure rotation results in angles near π/2; pure scaling/expansion results in angles near 0. Distributions include all analyzed times and conditions. Dots at top show distribution peaks for individual datasets. b. Quality of the fit (R²) provided by M_skew relative to an unconstrained M. We assessed fit quality for both the rank 6 matrices that capture dynamics in all 6 analyzed PCs/jPCs (bottom row) and the rank 2 matrices that capture dynamics in the first jPCA plane (top row). Circles plot performance for individual datasets. Squares give overall averages. Asterisks indicate a significant difference (t-test, p<0.05) from neural data. c. Average (across monkeys A and B) neural trajectory for all instructed-slow conditions (green) and all instructed-fast conditions (red). d. Similar to c but for the generator model.