Neuronal representation of saccadic error in macaque posterior parietal cortex (PPC)

Abstract

Motor control, motor learning, self-recognition, and spatial perception all critically depend on the comparison of motor intention to the actually executed movement. Despite our knowledge that the brainstem-cerebellum plays an important role in motor error detection and motor learning, the involvement of neocortex remains largely unclear. Here, we report the neuronal computation and representation of saccadic error in macaque posterior parietal cortex (PPC). Neurons with persistent pre- and post-saccadic response (PPS) represent the intended end-position of saccade; neurons with late post-saccadic response (LPS) represent the actual end-position of saccade. Remarkably, after the arrival of the LPS signal, the PPS neurons' activity becomes highly correlated with the discrepancy between intended and actual end-position, and with the probability of making secondary (corrective) saccades. Thus, this neuronal computation might underlie the formation of saccadic error signals in PPC for speeding up saccadic learning and leading the occurrence of secondary saccade.

Keywords: intention; motor error; neuroscience; non-human primates; proprioception; reference copy; rhesus macaque; saccadic eye movement.

Figure 1.. Conceptual model and behavioral paradigms.

(A) The proposed temporal sequence for comparison of the intended and the actual position signals. The intended signal raises up before the initiation of movement and lasts after the arrival of the external sensory signals. The external sensory signal raises up after the start of movement. The comparison occurs after the convergence of these two signals (marked by the dashed rectangle). The error signal is generated after the comparison. (B) Spatial-cue delayed saccade task (SCS). Horizontal direction trials: target is 10° in eccentricity. Oblique direction trials: target is 13° in eccentricity. (C) Memory-guided saccade task (MGS). (D) Color-cue delayed saccade task (CCS). DOI: http://dx.doi.org/10.7554/eLife.10912.002

Figure 2.. PPS neurons represent the intended end-position of saccades.

(A) The activity of an example PPS neuron in the SCS task. Averaged spike density with the 95% confidence interval (shaded area) in the preferred (black) and null direction (grey) are shown in the upper panel, whereas the horizontal and vertical eye traces are shown in the lower panel. (B) The averaged population activity of 89 PPS neurons (69 neurons from monkey B, 20 neurons from monkey DT). (C) Comparison of each neuron’s activity between trials that saccades directed to the left or right target. Symbols represent the activity of single neurons. The horizontal and vertical bars represent the standard error of the mean (SEM) of pre-saccadic activity (-200~0 ms from saccade start). The filled dots denote that the activity is significantly different between left and right saccades (p<0.05, Wilcoxon test). (D) The correlation between perisaccadic activities (-150~100 ms after saccade end) of single PPS neurons (n = 55) and the end-position of the horizontal saccades. Each dot represents the correlation coefficient value of a single neuron. The histograms in horizontal and vertical axis represent the distribution of correlation coefficient and p value, respectively. (m: mean, p: p value, n: number of neurons). DOI: http://dx.doi.org/10.7554/eLife.10912.003

Figure 2—figure supplement 1.. Correlation between single PPS neuron’s pre-saccadic and post-saccadic activity.

The filled bars indicate a significant correlation (p<0.05, t-test). The solid vertical line represents the mean value of the correlation coefficient. Data show a significant positive correlation in population level (r = 0.2151, p<0.001, two tailed t-test). DOI: http://dx.doi.org/10.7554/eLife.10912.004

Figure 2—figure supplement 2.. Correlation between the single PPS neuron’s perisaccadic activity and the end-position of saccade when monkey made oblique saccades.

(A) The correlation between perisaccadic activity (-150~100 ms after saccade end) of single PPS neuron (n = 24) and horizontal post-saccadic eye position. Each dot represents the value of correlation coefficient analysis of a single neuron. (B) The correlation between perisaccadic activity of single PPS neuron (n = 24) with vertical post-saccadic eye position. In panel (A) and (B), the horizontal and vertical histogram represents the distribution of the value of correlation coefficient and p value, respectively. DOI: http://dx.doi.org/10.7554/eLife.10912.005

Figure 2—figure supplement 3.. The post-saccadic activity of PPS neurons is not evoked by foveal stimulation.

(A) The averaged activity of the PPS neurons is aligned at saccade end in the SCS and at visual feedback onset in the MGS. The shaded area represents the 95% confidence interval. The first visual peak in MGS was evoked by peripheral visual target. The same PPS neurons did not respond to the foveal stimulation in MGS. (B) The comparison of a single neuron’s post-saccadic activity in SCS and foveal stimulation in MGS. Each symbol represents the activity of a neuron. The horizontal and vertical bars represent the standard error of the mean (SEM) of activity. The filled dots denote that the activity is significantly different between two tasks (p<0.05, Wilcoxon test). DOI: http://dx.doi.org/10.7554/eLife.10912.006

Figure 2—figure supplement 4.. PPS neurons show a similar firing pattern but different rate between tasks.

(A) An example neuron’s activity in SCS and MGS tasks. Activity was aligned at saccade end (vertical dashed line). (B) The population activity of 55 PPS neurons. The shaded areas represent the 95% confidence interval. (C) Comparisons of perisaccadic activities (150 ms before to 100 ms after saccade completion) between the SCS and MGS tasks. Most neurons show greater perisaccadic activity in SCS task than that in MGS task (filled symbols, p<0.05, Wilcoxon test). (D) Comparison of relative change of perisaccadic activity between the SCS and MGS tasks. For most neurons, the relative change of perisaccadic activity is similar between two tasks (p>0.05, Wilcoxon test; p=0.2887, in population level, paired t-test). DOI: http://dx.doi.org/10.7554/eLife.10912.007

Figure 3.. LPS neurons represent the actual end-position of saccades.

(A) The activity of an example LPS neuron in the SCS task. Averaged spike density with 95% confidence interval (shaded area) in the preferred (black) and null direction (grey) are shown in the upper panel, whereas the horizontal and vertical eye traces are shown in the lower panel. (B) The averaged population activity of 27 LPS neurons. (C) Comparison of single neuron’s activity between trials that saccades directed to the left or right target. Symbols represent the activity of signal neurons. The horizontal and vertical bars represent the standard error of the mean (SEM) of post-saccadic activity (25–225 ms after saccade onset). The filled dots denote that the activity is significantly different between left and right saccade (p<0.05, Wilcoxon test). (D) The correlation between postsaccadic activity (25~125 ms after saccade offset) of single LPS neurons (n = 22) with horizontal post-saccadic eye position when monkey made horizontal saccades. Each dot represents the correlation result of a single neuron. The horizontal and vertical histograms represent the distribution of correlation coefficient or p value, respectively. DOI: http://dx.doi.org/10.7554/eLife.10912.008

Figure 3—figure supplement 1.. LPS neurons discharged similarly in different tasks.

(A) An example and (B) 17 neurons’ average activity in the preferred direction in SCS and MGS tasks. (C) Comparisons of absolute post-saccadic activities between the SCS and MGS tasks. (D) Comparison of the relative change of post-saccadic activity between the SCS and MGS tasks. DOI: http://dx.doi.org/10.7554/eLife.10912.009

Figure 4.. The integration between PPS and LPS neurons and the correlation with the end-position of saccades.

(A) The temporal relationship between the activities of two types of neurons. The averaged activity with a 95% confidence interval of PPS neurons (black, n = 89) and LPS neurons (red, n = 27) was superimposed. The black and red vertical lines marked the time of the middle point of activity change for PPS neurons and LPS neurons, respectively. (B-C) The mean activity of LPS neurons (B) (n = 24) and PPS neurons (C) (n = 55) in three subsets of trials that were grouped based on the post-saccadic eye position in the horizontal meridian. The inserted histogram represents the distribution of post-saccadic eye position in the horizontal meridian. Colored curves represent the population activity of trials that have same color in the inserted histogram. The solid rectangle in (B) marks the postsaccadic period which was used for the analysis in (D), while the dashed and solid rectangles in (C) mark the perisaccadic internal used in (D) and post-subtraction interval used in (E), respectively. (D) The correlations of the late post-saccadic activity (25~125 ms after saccade offset) of LPS neurons (red) and perisaccadic activity (-150~100 ms relative to saccade offset) of PPS neurons (black) with post-saccadic eye position in the horizontal meridian. The colored dots represent the normalized activity of correlated trials for two types of neurons, respectively. The colored lines represent the regression linear fitting of correlated activity. The black and red bars in the bottom and top represent the trial number of each group. (E) The correlation between the post-subtraction activity (150~350 after saccade offset) of PPS neurons and the horizontal postsaccadic eye position. The larger the difference between the postsaccadic eye position and target location, the greater the post-subtracted activity in both undershoot and overshoot saccades. (F) The correlation between post-subtraction activity (150~350 ms after saccade offset) of single PPS neuron (n = 55) with saccadic error (distance between post-saccadic eye position and target position in horizontal meridian). Each dot represents the correlation result of a single neuron. The horizontal and vertical histogram represents the distribution of correlation coefficient or p value, respectively. (G) The post-subtraction activity fits well with the absolute difference between intended and real eye position signals. The blue and red lines represent the linear fitting of normalized perisaccadic activity of PPS neurons (blue, intended eye position signal) and post-saccadic activity of LPS neurons (red, actual end-position signal) with horizontal post-saccadic eye positions, respectively. The black lines represent the linear fitting of post-subtraction activity of PPS neurons with horizontal post-saccadic eye positions. The pink dashed line represents the regression fitting of the absolute subtraction result between the post-saccadic activity of LPS neurons and the perisaccadic response of PPS neurons. DOI: http://dx.doi.org/10.7554/eLife.10912.010

Figure 4—figure supplement 1.. Averaged Pearson correlation coefficient (CC) analysis between post-subtraction activity of PPS neurons (n = 55) and saccadic errors for horizontal saccades in SCS task.

A 100 ms window with a sliding-step of 10 ms was used to analyze the post-subtraction activity. Each data point represents the averaged CC value within a 100 ms window. Numbers on the x-axis represent the middle point of the sliding window. Asterisks indicate the follows: *p<0.05, **p<0.01, ***p<0.001 (Paired t-test). DOI: http://dx.doi.org/10.7554/eLife.10912.011

Figure 4—figure supplement 2.. Correlation between post-subtraction activity and oblique saccadic errors.

(A) The average activity of two subsets of trials of PPS neurons (n = 28). The inserted histogram represents the distribution of trials as a function of saccadic errors. (B) The correlation between a single PPS neuron’s post-subtraction activity and saccadic error. (C) The correlations between the normalized population post-subtraction activity (150~350 ms after saccade end) and saccadic error. (D) The correlation between the probability of making corrective saccades and the saccadic error. (E) The correlation between post-subtraction activity and the probability of making corrective saccades. DOI: http://dx.doi.org/10.7554/eLife.10912.012

Figure 4—figure supplement 3.. Correlation between post-subtraction activity and horizontal saccades in color-cue delayed saccades.

(A) The average activity of PPS neurons in two subsets of trials that were grouped based on the post-saccadic eye position in the horizontal meridian (n = 27). The inserted histogram represents the distribution of trials as a function of horizontal post-saccadic eye positions. Different colors represent the trials being classified in different groups. The rectangle marks the post-subtraction interval in which the activity was further analyzed. (B) The correlation between a single PPS neuron’s post-subtraction activity and saccadic error. Each dot represents a single neuron (n = 27). The histograms in horizontal and vertical axes represent the distribution of the value of correlation coefficient and p value, respectively. (C) The correlations between the normalized population post-subtraction activity (150~350 ms after saccade end) and saccadic error. The black line represents the linear regression fitting. The black bars below each data point represent the trial number of each group. (D) The correlation between the probability of making corrective saccades and the horizontal post-saccadic eye position. (E) The correlation between post-subtraction activity and the probability of making corrective saccades. DOI: http://dx.doi.org/10.7554/eLife.10912.013

Figure 5.. The correlation between the post-subtraction activity and the secondary saccades.

(A) The correlation between the probability of making corrective saccades and the post-saccadic eye position. Data showed the lower the saccadic accuracy, the higher the secondary saccade ratio. Two black lines represent the linear fittings of undershoot and overshoot saccades, respectively. (B) Correlation between the post-subtraction activity and the occurrence of a secondary saccade. The data groups were the same as in panel (A). (C) The average activity in trials with and without corrective saccades. Among trials with similar horizontal post-saccadic eye position, the post-subtraction activity was higher in trials with secondary saccades (solid curves) than in trials without secondary saccades (dashed curves). (D) The comparison of the activity between trials with and without secondary saccades. All data sets show that trials with secondary saccades have higher post-subtraction activity. The length of the bars under each data point represents the trial number in each group (red: trials with corrective saccade; black: trials without corrective saccade). Asterisks indicate the following: *p<0.05, **p<0.01, ***p<0.001 (Wilcoxon test). DOI: http://dx.doi.org/10.7554/eLife.10912.014

Figure 6.. The 3D distribution of the PPS and LPS neurons.

The distributions of two groups of neurons based on recording site are shown for both monkeys (A, B). Each circle represents one single neuron. The black and red stars mark the averaged center positions for PPS and LPS neurons, respectively. The mean distances between the PPS and LPS neurons are 1.78 mm and 1.0727 mm for monkey B and monkey D, respectively. The X and Y coordinates are relative to the center of the recording chamber. DOI: http://dx.doi.org/10.7554/eLife.10912.015

Author response image 1.. We used Gaussian mixture model (neglecting covariance) to classify all neurons (LPS+PPS) into two groups by using expectation maximization.

For each single neuron, the mean activity around the saccade (400 ms before, to 400 ms after saccade onset, time bin = 50 ms) was used in the analysis resulting in a 16-dim vector for each neuron. For simple illustration of the 16-dimensional space, we used linear Fisher’s discriminant dimension to project the space onto meaningful dimensions. The two clusters are shown in red and blue, respectively. The ellipses represent the standard deviation of two Gaussians. The clustering result fits our previous neuron grouping (LPS vs. PPS) very well, except that few PPS neurons (red dots, 7 neurons) were classified as in the same cluster as LPS neuron. DOI: http://dx.doi.org/10.7554/eLife.10912.016

Author response image 2.. The averaged population activity of 27 LPS neurons in the SCS task.

Averaged spike density with SEM (shaded area) in the preferred (black) and null direction (grey) are shown separately. DOI: http://dx.doi.org/10.7554/eLife.10912.017