Abstract

Visuomotor transformation is a fundamental process in executing voluntary actions. The final steps of this transformation are presumed to take place in the corticospinal (CS) system, yet the way in which the motor cortex (MC) interacts with spinal circuitry during this process is unclear. We studied neural correlates of visuomotor transformation in the MC and cervical spinal cord while monkeys performed an isometric wrist task. We recorded 2 measures of population activity: local field potential (LFP), reflecting local synaptic inputs and multi-unit activity (MUA), reflecting spiking activity emitted by nearby neurons. We found robust cortical and spinal responses locked to visual and motor events. In motor cortex, LFP responses were predominantly visually related; MUA responses were mostly motor related. Spinal LFP responses were generally weak, yet spinal MUAs showed visual and motor responses with distinctive patterns. For both structures, amplitudes of visual responses were positively correlated with amplitudes of motor responses and negatively correlated with reaction times. The temporal relations of cortical and spinal responses shifted from weak coactivation before movement to increased coupling following torque onset, with cortical leading spinal activity. Thus, ongoing CS interactions may exist at early stages of movement preparation. These interactions are dynamic and may shape the executed motor action.

Introduction

A critical step that precedes most volitional motor actions includes a visuomotor transformation (Caminiti et al. 1998) that involves translating the externally defined parameters of a visual target into a muscle-based motor command. However, the relative contribution of parallel versus serial processing in this transformation is still unknown. Classical views of the motor system emphasized the existence of distinct “high-level” and “low-level” functional modules with clear anatomical correlates (Wise 1985). Accordingly, it was suggested that information flows in a sequential top-down manner (Kalaska et al. 1997; Fagg and Arbib 1998) from posterior parietal areas through the motor cortex (MC), including the premotor (PM) cortex and the primary motor cortex (MI) to finally reach the spinal cord (SC). However, several electrophysiological studies have found nearly parallel activation of multiple cortical and subcortical motor areas (Kalaska et al. 1983; Crutcher and Alexander 1990; Ashe and Georgopoulos 1994) during preparation for an upcoming movement, consistent with a parallel mode of processing. These studies have typically used an instructed delay paradigm that forces a delay between a Cue signal indicating the required action, and a Go signal that instructs the subject to execute this action. This paradigm enables a temporal segregation of neural responses related to the sensorimotor transformation from responses related to movement execution and control.

Using an instructed delay task, it was found that spinal interneurons often modify their firing rate during the delay period in the absence of simultaneous changes in muscle activation (Prut and Fetz 1999). This finding contrasts with the common view of the SC in the context of voluntary movements as an output relay that merely executes preplanned motor commands. However, the data from this study did not allow for an analysis of cortical and spinal responses recorded simultaneously during the same task conditions.

The objective of the present study was to identify neuronal correlates of visuomotor processing in the corticospinal (CS) system and characterize CS interactions, using population activity measures. For this purpose, we simultaneously recorded spinal and motor cortical activity while monkeys performed a 2D isometric wrist task with an instructed delay period. We used 2 population-based signals, local field potential (LFP) and multi-unit activity (MUA). The LFP is considered to reflect the activity of synchronized synaptic inputs from a radius of up to a few mm (Mitzdorf 1985; Bullock 1997), whereas the MUA reflects spiking activity, that is, local output, from a radius on the order of 100 μm (Buchwald et al. 1965). Because previous findings indicated a widespread CS interaction during movement preparation and execution (Yanai et al. 2007), these interactions were expected to be optimally reflected in population measures. We found that cortical and spinal population signals expressed clear visual and motor responses, suggesting an early CS processing of motor-related parameters. Strong visual responses tended to predict stronger motor responses, as well as shorter reaction times. Cross-correlations of event-related cortical and spinal LFPs and MUAs were dynamic and consistent with a shift from a weak synchrony of the MC and the SC to a more tightly coupled “command mode” around movement execution.

Materials and Methods

All experimental procedures are described in detail elsewhere (Yanai et al. 2007, 2008). A brief description of the relevant procedures is given here.

Animals and Behavioral Task

Three monkeys (Maccaca fascicularis, females, 3–4 kg) performed an isometric 2D wrist task with an instructed delay period (Fig. 1A). During task performance, the monkey held its hand in either a pronation or supination position and controlled a cursor on a computer screen by applying a 2D isometric torque at the wrist (flexion/extension and radial/ulnar). A trial was initiated by the appearance of a central target (Trial onset). The monkey positioned the cursor inside the target by generating 0 torque for a rest period (500–600 ms). Then, 8 uniformly distributed peripheral targets appeared around the center target at a fixed distance, defining the onset of a delay period. One of these peripheral targets was presented in a distinct color for 500 ms (Target onset). The disappearance of the central target (monkey V, 1300–1700 ms, monkey D & G, 850–1200 ms after Target onset) served as a “Go” signal. The monkey then had to move the cursor into the previously cued target by generating an isometric torque in the appropriate direction and magnitude, and to keep the cursor within the target for an active torque period (350–750 ms). Subsequently, the peripheral targets disappeared, and the central target reappeared. The monkey returned to the rest position and received a reward, after which the screen went blank for 1000–1500 ms and a new trial started. The timing between externally induced events (e.g., between Cue onset and Go signal) was randomly varied from trial to trial.

Figure 1.

Experimental setup and recording sites. (A). Recording configuration and trial sequence. Monkeys sat in front of a computer screen and controlled an on-screen cursor by applying isometric 2D wrist torque. UNIT and LFP and recordings were made from MC (blue) and cervical SC (red) simultaneously. The compound single unit data were subsequently translated into MUA (see text). Bottom: An example of torque magnitude as a function of time during a single trial, Monkey V. The vertical ticks indicate the main behavioral events. Time at which target is filled (cue on) is marked by a blue rectangle. Torque and neural data shown in this example were recorded simultaneously. (B). Cortical (top) and spinal (bottom) penetration maps of the 3 monkeys. AS: arcuate sulcus, CeS: central sulcus. Initials C5 to T2 indicate vertebrae.

Figure 1.

Experimental setup and recording sites. (A). Recording configuration and trial sequence. Monkeys sat in front of a computer screen and controlled an on-screen cursor by applying isometric 2D wrist torque. UNIT and LFP and recordings were made from MC (blue) and cervical SC (red) simultaneously. The compound single unit data were subsequently translated into MUA (see text). Bottom: An example of torque magnitude as a function of time during a single trial, Monkey V. The vertical ticks indicate the main behavioral events. Time at which target is filled (cue on) is marked by a blue rectangle. Torque and neural data shown in this example were recorded simultaneously. (B). Cortical (top) and spinal (bottom) penetration maps of the 3 monkeys. AS: arcuate sulcus, CeS: central sulcus. Initials C5 to T2 indicate vertebrae.

Recording Techniques

After each monkey was fully trained, a piece of bone was removed above the frontal lobe and a cortical chamber (27 × 27 mm) was attached to the skull using orthopedic screws and dental acrylic. The surgical procedure was carried out in aseptic conditions, under general anesthesia (halothane anesthesia, induced by medetomidine hydrochloride [Domitor, 0.1 mg/kg] and ketamine [3 mg/kg]). Analgesia (carprofen [Rymadil]) and antibiotics (ceftriaxone) were administered peri-operatively. The dura mater was left intact. Arm-related sites and approximate boundaries of MI and dorsal PM cortex were mapped using trains of microstimulation pulses (50 ms of biphasic stimulation given at 300 Hz with intensity ≤ 60 μA). Subsequently, a laminectomy of the C5-T1 vertebrae was performed and a recording chamber (17 mm × 27 or 37 mm) was implanted over the lower cervical spine, above the cervical enlargement. Spinal recordings targeted spinal interneurons located at intermediate lamina that are the main target of CS inputs (Kuypers and Brinkman 1970; Bortoff and Strick 1993). Spinal and cortical LFPs were recorded using tungsten electrodes located deep in the gray matter in a differential mode of recordings relative to the reference signals obtained from the guide tubes (spinal and cortical, respectively) that were lowered to touch the dural tissue and a remotely located ground. Unit activity (300–6000 Hz) and LFP (1–150 Hz) were both collected from the same electrode by routing the raw signal to 2 amplifiers. LFP was amplified (gain = 5000) and digitized using sampling rate of 1 kHz. Wrist torque and behavioral events were recorded (1 kHz).

After recordings were completed, 2 of 3 monkeys (D and G) were deeply anesthetized (ketamine, pentobarbital, 30 mg/kg), and pins were inserted into known coordinates of the cortical implant. The animals were then euthanized with pentobarbital sodium (50 mg/kg IV) and perfused with 10% Formalin or 4% paraformaldehyde. Cortical locations of penetrations relative to anatomical landmarks were subsequently reconstructed (Fig. 1B). Monkey V was sent to a primate sanctuary, after removal of the recording chamber and suturing of the scalp skin. Reconstructions of recording sites in this monkey were based on aligning recording coordinates with surface landmarks. All surgical and animal handling procedures were according to the NIH Guide for the Care and Use of Laboratory Animals (1996), complied with Israeli law, and were approved by the Ethics Committee of the Hebrew University. Animal care was supervised by the veterinarian staff of the Hebrew University.

Data Analysis

LFP signals were preprocessed offline by removal of 50 Hz line-related noise, band-pass filtering of 3–100 Hz (Butterworth, 3 pole), and downsampling to 500 Hz. Line-induced noise was removed using an adaptive filter method (Schoffelen et al. 2005), with a window of ±2 bins (0.5 Hz) to compensate for phase variability of the AC line encountered in our data. Next, event-related LFPs (eLFPs) were computed by first aligning all trials on any of the 3 behavioral events: Trial onset, Target onset or Torque onset, and then averaging across trials. The timing of Torque onset was determined offline by an ad hoc method based on a threshold crossing of the time derivative of the exerted torque. Specifically, torque data were smoothed (using a Gaussian of 20 ms standard deviation, SD) and then derived. The time of maximal derivative (between GO signal and Target acquisition) was first identified. Torque onset was defined as the earliest time (before the time of the maximal derivative) at which torque-derivative was lower than a threshold level. The actual threshold differed slightly across monkeys, due to differences in task performance. After applying the algorithm, we visually inspected single trials to avoid any erratic identification of torque onset made during the automatic processing.

The eLFPs were computed separately for each behavioral condition (8 target directions × 2 hand positions) as well as for data pooled over all behavioral conditions. However, to enhance signal-to-noise ratios, this report includes only analyses of eLFPs pooled over target directions and hand position (but computed separately for each of the 3 behavioral events). This analysis produced a single averaged response for each site and behavioral event. To overcome variations in signal amplitude across recording sites, the eLFPs of each site were standardized. In this manner, it was possible to directly compare event-related responses that were obtained for different behavioral events. Background mean level and SD were calculated for the 500 ms epoch before Trial onset, a period where no task-related modulation was expected. This epoch contained no torque change, because the monkeys produce 0 torque throughout the intertrial interval. Next, each of the 3 eLFPs (Trial onset, Target onset, and Torque onset) was standardized by subtracting the mean baseline level and dividing by the SD of the baseline epoch. Significant eLFPs were identified as those deviating significantly from the ±3SD confidence interval for at least 10 ms.

MUA was computed using the procedure described by Gail et al. (2004). The raw unit activity signal was subjected to a band-pass filter (300–6000 Hz), full wave rectified, and low-pass filtered with a cutoff frequency of 100 Hz. The mean prestimulus signal was subtracted from task epochs and traces were cut, aligned, and averaged as was done for the LFP signal, to obtain event-related MUAs (eMUAs).

Timing of Event-Related LFPs and MUAs

We estimated the timing of LFP and MUA responses to the 3 different events (Fig. 2). Algorithms that were designed for detecting onset time of noisy signals often provide measures that are sensitive to the signal-to-noise ratio of the tested signals. In our data, we could not use this approach. To overcome this problem, we quantified the timing of each response as the time of peak activity. However, as the temporal properties and typical response pattern of eMUA and eLFP differed greatly the applied definition of “peak activity” was somewhat different for these 2 signals. ELFPs typically showed multiple, positive, and negative peaks of varying amplitude (Fig. 2A–C), whereas eMUAs showed a typical monotonic excitation, sometimes preceded by a brief inhibition (Fig. 2D–F). Therefore, peak LFP activity was defined as the maximal absolute peak over the epoch −250 to +750 ms relative to each event, whereas peak MUA activity was defined as the maximum absolute derivative (signal smoothed with a Gaussian window with SD = 40 ms) for this epoch.

Figure 2.

Estimating timing of LFP and MUA responses. Determining the timing of responses of eLFP (AC) and eMUAs (D–F). (A) average LFP around target onset computed across all sessions for a single monkey (V). (B) Four examples of single-site significant LFP responses obtained for the same event and monkey as in (A). For each trace, arrows indicate the timing that was defined for that response. Note that the relative size of the different components of the evoked response vary substantially across the different sites. (C) Distribution of timing for all significant responses for target onset in monkey V. Mean response time is marked with gray triangles. (D) Average MUA around target onset across all sessions for monkey V. (E) Single-site eMUA (gray) and the signal processing applied to determine response timing: smoothed (dashed line) and derived (black line). The derivative here is shown in a different scale than the raw signal for clarity. The maximal derivative (black dot) was taken as the response timing for that site. (F) Distribution of MUA response timing for all significant responses to target onset in monkey V. Gray triangles mark the average response time.

Figure 2.

Estimating timing of LFP and MUA responses. Determining the timing of responses of eLFP (AC) and eMUAs (D–F). (A) average LFP around target onset computed across all sessions for a single monkey (V). (B) Four examples of single-site significant LFP responses obtained for the same event and monkey as in (A). For each trace, arrows indicate the timing that was defined for that response. Note that the relative size of the different components of the evoked response vary substantially across the different sites. (C) Distribution of timing for all significant responses for target onset in monkey V. Mean response time is marked with gray triangles. (D) Average MUA around target onset across all sessions for monkey V. (E) Single-site eMUA (gray) and the signal processing applied to determine response timing: smoothed (dashed line) and derived (black line). The derivative here is shown in a different scale than the raw signal for clarity. The maximal derivative (black dot) was taken as the response timing for that site. (F) Distribution of MUA response timing for all significant responses to target onset in monkey V. Gray triangles mark the average response time.

The approach we used for timing eLFPs is somewhat different than previous reports (e.g., N1; Donchin et al. 2001) where the timing of a specific peak (e.g., the first negative peak-N1) was used for all sessions. Our data were characterized by considerable across-session variability so that the initial peak of the response was not necessarily of the same polarity (Fig. 2B). This variability in the eLFP between sessions led us to adopt an approach that was independent of any specific order of peaks (negative or positive) in the eLFP. Although this approach sometimes overestimated the timing of the eLFP, it was robust for responses triggered by different events (visual and motor) and at different sites (cortical and spinal). Note that previous reports of specific eLFP peaks (e.g., Donchin et al. 2001; Roux et al. 2006) focused on “movement”-related cortical LFPs. Comparing our method with a more restrictive search for specific peaks within a given time window (selected based on the average response) yielded very similar results.

In contrast to eLFPs, eMUAs typically lacked a clear peak, indicating maximal population firing rate, especially when the data were aligned to visual events (Fig. 2D). Thus, defining the time of MUA peak activity based on the timing of the maximum value (or maximal absolute value) was very sensitive to noise and tended to produce biased timing estimates (Supplemental Fig. 1). We thus focused on times of maximal “change” in population rates, using the derivative of the smoothed eMUA (Fig. 2E). This approach resulted in unimodal timing distributions of eMUAs (Fig. 2F).

In order to quantify the temporal interrelations of eLFPs/eMUAs, cross-correlations were computed. For this purpose, spinal and cortical eLFPs and eMUAs were recomputed using alternate trials, to cancel out potential effects of electrical cross-talk residuals. Each eLFP/eMUA was computed for a 1000-ms epoch starting 500 ms before the event. Raw cross-correlations were computed (maximum lag: ±400 ms) and normalized such that the peak of the autocorrelation equaled 1 (and therefore the cross-correlation at 0 lag equaled the correlation coefficient), and to compensate for the different data sizes available for each time lag (Kohn and Smith 2005). Correlation amplitude was estimated by the global maximum over the ±400 ms lag range. We also computed cross-correlations and measured their peaks in the same manner separately for the 2 narrower time windows: the 500-ms epoch before each event and the 500-ms epoch after it. By computing the 2 correlations separately for data recorded before and after events, we were able to compare the impact of different behavioral events on the eLFP and eMUA interactions.

Results

We studied a total of 279 pairs of spinal and cortical LFPs and MUAs, recorded in 71 recording sessions (Table 1). Figure 1B shows the locations of spinal and cortical penetrations, and the representation of arm and hand-related sites on the cortical surface, as determined by microstimulation. Only arm-related sites were included in the analyses. The boundary between PM and M1 was determined using microstimulation thresholds. Sites in which responses were obtained above 15-μA stimulation were classified as PM. Using these criteria, we were able to draw a rough boundary between M1 and MP so that mismatching sites (high threshold but near the central sulcus or low threshold near the arcuate sulcus were removed from the analysis. Findings for MI sites and PM sites were generally consistent, and hence unless stated otherwise, are grouped under the common term “motor cortex.”

Table 1

Numbers of recording sites, by monkey and recording area

  Monkey V Monkey D Monkey G Total 
MC MI 59 15 26 100 
PM 37 40 
Total 96 16 28 140 
SC 87 23 29 139 
  Monkey V Monkey D Monkey G Total 
MC MI 59 15 26 100 
PM 37 40 
Total 96 16 28 140 
SC 87 23 29 139 

Early Visual Responses in Cortical LFP and MUA

Event-related responses to visual events were observed for both LFPs and MUAs following the Trial onset and Target onset events (Fig. 3). In some cases, aligning the LFP traces on Trial Onset revealed a nonsignificant response (Fig. 3A) compared with the significant response when the same traces were aligned on Target Onset (Fig. 3B), yet both averages were similar in shape. The main differences were smaller peak amplitudes and a slight delay in the timing of each peak (e.g., the latency of the strongest negative peak is >200 ms in Fig. 3A and <200 ms in Fig. 3B). An explanation relating these eLFPs and eMUAs to a motor response could be ruled out, because neither visual event produced significant changes in torque magnitude in any of the monkeys (compare Fig. 3D,E with Fig. 3F).

Figure 3.

Examples of cortical visual and motor responses. (AC) Examples of eLFP (blue) and eMUA (red), observed in a single site from monkey V. The same traces were aligned on the Trial onset event (A), Target Onset event (B), and Torque onset (C) and then averaged. Dashed lines denote ±3SD lines. (DF) Averages of torque amplitude, computed by aligning the torque traces on the same behavioral events. Traces show the average torque for each monkey (V, D,G) obtained by averaging for each site and then averaging across sites. Purple line (V*) is the average torque trace for a single site corresponding to the site from which the examples in (AC) were taken.

Figure 3.

Examples of cortical visual and motor responses. (AC) Examples of eLFP (blue) and eMUA (red), observed in a single site from monkey V. The same traces were aligned on the Trial onset event (A), Target Onset event (B), and Torque onset (C) and then averaged. Dashed lines denote ±3SD lines. (DF) Averages of torque amplitude, computed by aligning the torque traces on the same behavioral events. Traces show the average torque for each monkey (V, D,G) obtained by averaging for each site and then averaging across sites. Purple line (V*) is the average torque trace for a single site corresponding to the site from which the examples in (AC) were taken.

The eLFPs triggered by Target Onset in MC were highly stereotypic in shape with little variance in peak latency across sites in any of the monkeys and were characterized by multiple peaks, the first of which occurred ∼100 ms after Target onset, and the last about 400 ms later (Fig. 4A). By contrast, significant eMUAs following Target onset typically indicated a gradual increase in population firing rate that leveled off at around 300–400 ms after stimulus onset and remained elevated throughout the delay (Fig. 4B). Cortical responses to Target onset were overall stronger (Mann–Whitney test, LFP: P < 10−4, MUA: P < 10−4) and more common (exact binomial test, LFP: P < 10−21, MUA: P < 10−7) than responses to Trial onset (Table 2). The prominence of responses to Target onset compared with Trial onset was corroborated by the slightly earlier eMUA responses to the former event (medians: 243 vs. 362 ms, P < 0.01, Mann–Whitney test). This was not the case for eLFPs (medians: 274 vs. 268 ms, P = 0.95). The timing distributions of cortical eLFP and eMUA responses to Target onset displayed an almost perfect overlap (Fig. 4C), but note that the timing data were computed using different algorithms for each signal (see Methods). Also note that there were a few cases in which a response was timed by our algorithm as preceding the visual event because the search window included a 250-ms preevent epoch. Clearly, these rare cases represent weak responses with a relatively low signal-to-noise ratio, which were falsely identified by the algorithm.

Table 2

Numbers (and percentages) of sites with significant evoked response following each behavioral event

Recording area Neural signal Trial onset Target onset Torque onset 
MC: N = 140 LFP 14 (10.0%) 93 (66.4%) 58 (41.4%) 
MI 8 (8%) 67 (67%) 35 (35%) 
PM 6 (15%) 26 (65%) 23 (57.5%) 
MUA 41 (29.3%) 86 (61.4%) 114 (81.4%) 
MI 26 (26%) 58 (58%) 86 (86%) 
PM 15 (37.5%) 28 (70%) 28 (70%) 
SC: N = 139 LFP 12 (8.6%) 12 (8.6%) 17 (12.2%) 
MUA 83 (59.7%) 88 (63.3%) 119 (85.6%) 
Recording area Neural signal Trial onset Target onset Torque onset 
MC: N = 140 LFP 14 (10.0%) 93 (66.4%) 58 (41.4%) 
MI 8 (8%) 67 (67%) 35 (35%) 
PM 6 (15%) 26 (65%) 23 (57.5%) 
MUA 41 (29.3%) 86 (61.4%) 114 (81.4%) 
MI 26 (26%) 58 (58%) 86 (86%) 
PM 15 (37.5%) 28 (70%) 28 (70%) 
SC: N = 139 LFP 12 (8.6%) 12 (8.6%) 17 (12.2%) 
MUA 83 (59.7%) 88 (63.3%) 119 (85.6%) 
Figure 4.

Cortical and Spinal responses to Target onset. (A). Grand averages of significant eLFPs across all cortical sites, computed separately for each monkey. Traces were aligned to Target Onset event. Average responses in each site were standardized (see Methods) to overcome variance in signal amplitude across sites. Note that the average torque trace around this event is shown in Figure 3E. (B) Grand averages of eMUAs for the same behavioral event. (C) Distributions of cortical eMUA and eLFP latencies of maximal activity (see text for details). The dashed lines indicate medians. (D) Grand averages of significant spinal eLFPs. (E) Grand averages of spinal eMUAs. Inset: grand average of spinal eMUAs from the 3 monkeys (black), superimposed on the grand average of cortical eMUAs (gray). For illustrative purposes, eMUAs were aligned to 0 amplitude at time of Target onset. (F). Distributions of spinal eLFP and eMUA latencies of maximal activity.

Figure 4.

Cortical and Spinal responses to Target onset. (A). Grand averages of significant eLFPs across all cortical sites, computed separately for each monkey. Traces were aligned to Target Onset event. Average responses in each site were standardized (see Methods) to overcome variance in signal amplitude across sites. Note that the average torque trace around this event is shown in Figure 3E. (B) Grand averages of eMUAs for the same behavioral event. (C) Distributions of cortical eMUA and eLFP latencies of maximal activity (see text for details). The dashed lines indicate medians. (D) Grand averages of significant spinal eLFPs. (E) Grand averages of spinal eMUAs. Inset: grand average of spinal eMUAs from the 3 monkeys (black), superimposed on the grand average of cortical eMUAs (gray). For illustrative purposes, eMUAs were aligned to 0 amplitude at time of Target onset. (F). Distributions of spinal eLFP and eMUA latencies of maximal activity.

Early Visual Responses in Spinal LFP and MUA

Significant spinal responses to both visual events were observed (Fig. 4D–F, LFP, monkeys V & G; MUA, all 3 monkeys, data for Trial onset not shown). LFP responses to Target onset were similar in amplitude to responses to Trial onset (Mann–Whitney test, P = 0.64), yet MUA responses to Target onset were stronger than responses to Trial onset (P < 10−2), as was observed in the cortical sites. However, the relative frequency of significant responses of each signal to the 2 events was similar (Table 2, exact binomial test, LFP: P = 1, MUA: P = 0.54). Peak activation times of spinal eMUAs and eLFPs were not significantly different, nor were differences between responses to Trial onset and Target onset (2-way Kruskal–Wallis test).

A grand average of the spinal eLFPs following Target onset revealed a significant complex wave with a latency that was similar to that of the first peaks of the cortical grand average (Fig. 4D). However, in contrast to the typical monotonic activation observed in cortical MUAs, spinal MUAs indicated a brief “decrease” in population firing ∼150 ms after Target onset, followed by gradual increased activity (Fig. 4E). The latter increase was similar in shape and timing to the response observed in MC. Such a biphasic average response may reflect a mixture of sites that produce monophasic excitation or monophasic inhibition at different times relative to Target onset. We therefore categorized the single-site eMUAs and found that the 2 most common response patterns were an “inhibition–excitation” pattern, similar to that shown in the grand means, and a monophasic excitation response. In contrast, monophasic inhibition was relatively rare among significant spinal eMUAs computed for Target onset (4/88 spinal sites).

Movement-Related Responses in Cortical and Spinal Sites

Motor eLFPs and eMUAs were common in MC (Table 2). Nevertheless, aligning cortical LFPs on Torque onset produced “fewer” significant responses than aligning the same data on Target onset (58/140 vs. 93/140 sites, respectively, P < 10−4). Moreover, significant motor eLFPs were observed only in 2 of the 3 monkeys (V and G) unlike visual responses that were observed in all monkeys. Note that monkey D had a slightly slower movement time, which may explain the lack of motor eLFPs. In contrast, significant motor eMUAs were more common than visual eMUAs related to Target onset (114/140 vs. 86/140, P < 10−3). We found no consistent differences in the relative frequency of either visual or motor responses between MI and PM sites.

The shape of the grand average motor eLFP to torque onset (Fig. 5A) was consistent with the P1–N1–P2–N2 waveform reported by Donchin et al. (2001). However, in previous reports of eLFPs, during a reaching task (Donchin et al. 2001; Roux et al. 2006), the first peak (“P1”) appeared prior to movement onset, whereas in our data, the first peak appeared following torque onset in the grand average eLFP and premovement peaks were uncommon in single-site data. This discrepancy may be due to task differences, because whole-arm reaching involves the recruitment of larger neural populations. Also, our definition of torque onset may provide an earlier measure than movement onset. It is impossible to determine whether the previously reported premovement “P1” was delayed such that it appeared following torque onset or that it was simply too weak to rise above the noise level.

Figure 5.

Cortical and Spinal motor responses. Traces were aligned to the Torque Onset event. All conventions are as in Figure 4.

Figure 5.

Cortical and Spinal motor responses. Traces were aligned to the Torque Onset event. All conventions are as in Figure 4.

The waveforms of motor eLFP in MC were similar to those of visual eLFPs, with the exception that visual eLFPs tended to have sharper (i.e., short duration) peaks (compare Figs. 3B and 4A with Figs. 3C and 5A). Grand averages of significant eMUAs showed a phasic increase in population firing (reaching a peak ∼150 ms after Torque onset) with a somewhat slower response in monkey D (Fig. 5B). The onset of motor eLFP lagged behind the observed motor eMUA by almost 100 ms (Fig. 5C, median values: 42 vs. 138 ms after Torque onset, Mann–Whitney test, P < 10−6), possibly reflecting the temporary dominance of sensory feedback induced by the action itself.

Spinal eLFPs and eMUAs also showed motor responses (Fig. 5D–F; LFP, monkeys V & G; MUA, all 3 monkeys) that differed from cortical responses in several respects. First, motor eLFPs in SC were less frequent than in the MC (exact binomial test, P < 10−7) and did not express a consistent and characteristic shape as observed in the cortical motor eLFPs. In fact, the distribution of latencies of these responses (Fig. 5F) lacked any prominent mode. In contrast, spinal motor eMUAs were robust and as frequent as cortical responses (exact binomial test, P = 0.35). Grand averages of the significant eMUAs showed that population activity in SC increased around Torque onset, reached a peak about 200–400 ms afterward, and remained elevated throughout the torque ramp (compare with the average torque magnitudes in Fig. 3F). This feature of spinal eMUAs, especially when contrasted with cortical eMUAs (compare Fig. 5B and Fig. 5E), highlights their active role in maintaining constant muscle force despite reduced cortical inputs.

We examined the relations between visual and motor eLFPs (or eMUAs) by computing the Pearson correlation coefficient between peak amplitudes at Target onset and peak amplitudes at Torque onset. Overall, strong visual responses tended to predict strong motor responses (Fig. 6A–D). Moreover, we found a negative correlation between peak amplitudes of the visual responses and monkey reaction times, defined as the interval between the Go signal and Torque onset (Fig. 6E–H). Thus, traces of behavioral efficiency could be seen at early stages of responses to visual cues by the cortical and spinal populations.

Figure 6.

Correlations between visual responses, motor responses, and “behavior.” (AD) Scatter plots display the association between standardized amplitudes of the visual response and the motor response, for each signal (LFP and MUA) and recording location (MC and SC). (EH). Scatter plots display the association between amplitudes of the visual response and average reaction time per recording session. Significance of linear correlation coefficients was tested using a parametric test and verified by a bootstrap method.

Figure 6.

Correlations between visual responses, motor responses, and “behavior.” (AD) Scatter plots display the association between standardized amplitudes of the visual response and the motor response, for each signal (LFP and MUA) and recording location (MC and SC). (EH). Scatter plots display the association between amplitudes of the visual response and average reaction time per recording session. Significance of linear correlation coefficients was tested using a parametric test and verified by a bootstrap method.

Dynamics of CS Interactions—Evidence from LFP Correlations

We studied the temporal relations between cortical and spinal eLFPs by computing the cross-correlation of eLFP pairs (Fig. 7A–C, Supplemental Fig. 2A–C). The shapes of the correlations of eLFP pairs were not consistent across subjects and recording sites. This may have been due to the low incidence of significant spinal eLFPs and their overall low signal-to-noise ratio. The median correlation peak lags following the 2 visual events (Fig. 7D,E) were not significantly different from 0 (−6 and −21 ms; Wilcoxon signed-rank test for 0 median: P = 0.97, 0.59, respectively), whereas the median correlation peak lag following Torque onset (Fig. 7F) was 58 ms (P = 0.003). This result is consistent with a cortex- to- spinal flow of information. The amplitude of the eLFP pairwise correlation tended to increase during a trial (Fig. 7G–I, Kruskal–Wallis test, P < 10−7). In order to examine whether changes in CS correlation amplitude were related to behavioral events, we recomputed the cross-correlation of eLFP pairs for 2 nonoverlapping time windows—before the event and after the event (see Methods). Then, for each pair of signals (cortical and spinal eLFPs), we compared the correlation found in the postevent data with the correlation computed for the preevent data. This analysis revealed that correlation amplitudes significantly decreased following Trial onset, as evidenced by the tendency of data points to concentrate below the line x = y (Fig. 8A; paired Wilcoxon signed-rank test). On the other hand, postevent correlation amplitudes significantly increased following Target onset and Torque onset (Fig. 8B,C) as demonstrated by the tendency of the data to concentrate above the x = y line. This increase in correlation amplitude suggests a tighter temporal locking between cortical and spinal populations.

Figure 7.

Cross-correlations between cortical and spinal eLFPs. (AC). Examples of cross-correlations of eLFPs computed for 1 site and shown separately for each behavioral event. (DF). Distributions of time lags of cross-correlation peaks computed for all possible pairs of recording sites. By convention, positive lags indicate MC leading over SC, negative lags indicate SC leading over MC. (G–I) Distributions of cross-correlation peak amplitudes. The dashed lines indicate medians.

Figure 7.

Cross-correlations between cortical and spinal eLFPs. (AC). Examples of cross-correlations of eLFPs computed for 1 site and shown separately for each behavioral event. (DF). Distributions of time lags of cross-correlation peaks computed for all possible pairs of recording sites. By convention, positive lags indicate MC leading over SC, negative lags indicate SC leading over MC. (G–I) Distributions of cross-correlation peak amplitudes. The dashed lines indicate medians.

Figure 8.

Cross-correlations of eLFPs, within-site comparison. (A–C). Scatter plots of eLFP cross-correlation peak amplitudes, plotted separately for each behavioral event. Asterisks mark pairs where the spinal eLFP was statistically significant (see Figs. 4 and 5). The dashed lines indicate medians of marginal distributions.

Figure 8.

Cross-correlations of eLFPs, within-site comparison. (A–C). Scatter plots of eLFP cross-correlation peak amplitudes, plotted separately for each behavioral event. Asterisks mark pairs where the spinal eLFP was statistically significant (see Figs. 4 and 5). The dashed lines indicate medians of marginal distributions.

Dynamics of CS Interactions—Evidence from MUA Correlations

Repeating the same analyses for cortical and spinal eMUAs revealed a generally similar picture to that observed in LFP correlations. Around Trial onset (Fig. 9A,D) and Target onset (Fig. 9B,E), eMUAs of SC and MC were weakly correlated but synchronized, as seen in the mode around 0 lag (median peak lags: 21, −7 ms, P = 0.54, 0.95, respectively). However, around Torque onset, the time lags of correlation peaks shifted toward positive values (median: 142 ms, P < 10−11), consistent with a dominance of motor cortical input to SC (Fig. 9C,F) accompanied by a relative decrease in the common input to both sites. The mode at 0 lag for correlations around visual events characterized only MI-SC cross-correlations, but not those of PM–SC site pairs, which did not show any prominent mode (data not shown). The grand mean of the eMUA correlations is shown in Supplemental Figure 2C,D.

Figure 9.

Cross-correlations between cortical and spinal eMUAs. (A–C) An example of cross-correlations of eMUAs computed separately for each behavioral event using data collected from a single site. (D–F) Distributions of time lags of cross-correlation peaks. (G–I). Distributions of cross-correlation peak amplitudes. All conventions are as in Figure 7.

Figure 9.

Cross-correlations between cortical and spinal eMUAs. (A–C) An example of cross-correlations of eMUAs computed separately for each behavioral event using data collected from a single site. (D–F) Distributions of time lags of cross-correlation peaks. (G–I). Distributions of cross-correlation peak amplitudes. All conventions are as in Figure 7.

The right shift in correlation peak times was accompanied by an increase in amplitude of eMUA cross-correlations (Fig. 9G–I; P < 10−15). Studying the peri-event modulation in correlation amplitude revealed no significant difference for data collected before and after Trial onset (Fig. 10A). In contrast, after both Target onset (Fig. 10B) and torque onset (Fig. 10C), a significant increase in correlation amplitude was observed.

Figure 10.

CS cross-correlations of eMUAs, within-site comparison. (AC). Scatter plots of eMUA cross-correlation peak amplitudes, plotted separately for each behavioral event. All conventions are as in Figure 8.

Figure 10.

CS cross-correlations of eMUAs, within-site comparison. (AC). Scatter plots of eMUA cross-correlation peak amplitudes, plotted separately for each behavioral event. All conventions are as in Figure 8.

The cross-correlation effects were only weakly related to the occurrence of evoked responses in single sites. We repeated the analysis using only those pairs with a significant spinal response and for pairs where both spinal and cortical responses were significant, and the results were qualitatively similar. The only difference was that using the statistical significance thresholds, the median correlation amplitudes increased, but the shift in peak lag was essentially the same. In summary, eLFP and eMUA cross-correlations displayed congruent effects that were independent of the amplitudes of each eLFP or eMUA.

Discussion

The objective of this study was to characterize visual and motor event–related activity of neural populations in the MC and the SC. We found robust responses in MC evoked by strictly visual events (Figs. 3 and 4). Such responses were found for signals reflecting both input to and output from MC. Furthermore, visual responses were also found in the cervical SC (Fig. 4D–F) at a frequency that at least for MUA, resembled the incidence of the motor cortical response to visual events. Robust motor eMUAs were observed in both sites (Fig. 5) reflecting their intimate involvement in the control of movement execution, whereas eLFPs were prominent solely in MC. The amplitudes of visual responses showed a moderate positive correlation with amplitudes of motor responses, and a weaker negative correlation with reaction time (Fig. 6). Cross-correlations of eLFPs and eMUAs indicated a functional coupling between MC and SC during movement preparation and execution that was dynamic and time locked to the course of the behavioral task (Figs. 7–10). Cross-correlations of eLFP and eMUA pairs were both consistent with a shift from a “coactivation mode” of MC and SC following visual events to a “command mode” around movement execution. This shift involved increased coupling magnitude as well as a change in relative timing, such that MC activity led spinal activity. This ongoing CS coupling provides a link through which early information processing of a planned motor action can gain access to low-level motor circuits and thus modify various parameters of muscle activity.

Common Features of Cortical and Spinal Visual Responses

Premovement tuning of single neurons or neuronal populations to informative cues of ensued movements (e.g., direction or extent) is often interpreted as evidence of early information processing that could facilitate behavior (e.g., Alexander and Crutcher 1990; Prut and Fetz 1999; O'Leary and Hatsopoulos 2006). Our findings provide additional evidence for the functional significance of visual responses in MC and SC. First, we found that visual responses in MC and SC to Target onset tended to be stronger and more frequent than responses to Trial onset. This result can be explained by differences in the quality of the visual cue (Fig. 1A). However, the 2 events differed primarily in their movement-related information content: the Target Onset cue contained a critical directional information imperative for task performance, whereas the Trial onset cue contained no directional information. This interpretation is consistent with findings relating the amplitude of the eLFP with the degree of uncertainty in the stimulus (Roux et al. 2006).

Second, we found a consistent positive correlation between population response amplitudes following the visual event (Target onset) and the motor event (Torque onset): Strong visual responses tended to predict strong motor responses. This correlation was significant in cortical eMUAs and eLFPs, as well as in spinal eMUAs. The correlation between visual responses and motor responses suggests that these attributes of motor action are not processed by distinct groups of neurons; rather, they are represented by a single group that presumably plays a role in visuomotor transformation.

Finally, the amplitudes of visual responses tended to be negatively correlated with the average reaction time per experimental session. These findings are in line with previous reports that have shown a correlation between delay-period activity and behavioral parameters such as reaction time and movement time (Alexander and Crutcher 1990; Riehle and Requin 1993; Churchland et al. 2006). Here, we show that such a behavioral correlation is already expressed at early stages of the delay-period activity and that this correlation is also found in spinal circuitry, which may thus serve as a link between early signal-related neural activity and the observed motor efficiency.

Visual Responses in Motor Cortex as Indicators of Visuomotor Transformation

Numerous previous studies have found visual responses in MC using single unit activity (e.g., Shen and Alexander 1997), LFP (Rubino et al. 2006), and EEG (Minelli et al. 2007). In addition, anatomical reports have documented the abundance of bidirectional pathways between frontal and parietal cortical areas (Luppino and Rizzolatti 2000). These findings conflict with earlier hypotheses that postulated a strict division of labor between the motor and sensory areas. Instead, sensorimotor areas are currently considered to participate in recurrent perception-action loops (Cisek 2007) where multiple areas contribute to both perceptual and motor processes (Lebedev and Wise 2002). Our findings are consistent with this view, given the robust visual responses of both eLFPs and eMUAs observed in MC, as well as the overall similarity of responses of MI and PM sites.

Moreover, we show that in the motor cortical sites examined, significant eLFPs (reflecting local input) were more frequently found in relation to visual events than to motor actions. This finding may be due to our use of a single-joint isometric task, during which cortical activation was possibly restricted compared with the extent of activation during whole-arm reaching and grasping. Alternatively, it is possible that in the context of the present task, cortical inputs related to behaviorally relevant visual events tended to be more synchronized or spatially organized than inputs related to motor events, which are essentially a mixture of motor, tactile, and proprioceptive inputs. Indirect support for this explanation comes from the fact that cortical eMUAs indicating local outputs from MC showed an opposite response pattern to eLFPs, namely, they were more “motor” than “visual.” This discrepancy between motor cortical input and its output may reflect a step in information processing within the MC during the delay period that is intimately related to visuomotor transformation. This interpretation is consistent with previous views suggesting that the MC is actively involved in translating spatial (i.e., visual in this case) signals into motor commands (Mushiake et al. 1997; Kakei et al. 1999, 2001; Yanai et al. 2008).

Biphasic Activity Changes Related to Visual Input in Spinal Neuronal Populations

It was previously reported that spinal reflexes (including monosynaptic ones) are modified prior to actual movements (Brunia et al. 1982; Komiyama and Tanaka 1990; Bonnet et al. 1991). Furthermore, spinal interneurons showed robust modulation during preparation for voluntary movement (Prut and Fetz 1999). This modulation is comparable with the delay-period activity reported for MI, PM cortex, supplementary MC, and the striatum (Tanji and Kurata 1985; Alexander 1987; Alexander and Crutcher 1990). Here, we found short-latency responses of spinal eLFPs and to a greater extent spinal eMUA, to visual events, where responses to Target onset were significantly stronger than responses to Trial onset. The 2 visual events signaled a forthcoming movement but differed first in their temporal proximity to the ensued motor action and second in their information content regarding the details of the action to be executed. The differential spinal response to the 2 events may thus suggest that spinal circuitry is informed regarding parameters of ensued movements very soon after this information is provided to supraspinal motor areas. Nonetheless, at this stage, we cannot rule out the possibility that spinal responses to visual events also include a more general readiness component.

The MUA responses to target onset consisted of an early inhibition that was locked to the event followed by a gradual excitation. This response pattern may reflect a summation of a mixed population of neurons, some of which express early inhibition whereas others respond with a later excitation, as was indeed reported for single interneurons (Prut and Fetz 1999). Note that this biphasic pattern deviated from the monophasic excitation exhibited by cortical eMUAs to target onset. The functional significance of this biphasic spinal response is yet to be determined. One possibility is that it reflects a cortical-derived “braking and priming” signal (Moll and Kuypers 1977; Sawaguchi et al. 1996; Prut and Fetz 1999), which enables preparation for the forthcoming motor action while avoiding its premature execution.

CS Interactions Dynamics Revealed by Population Measures

In a previous study that concentrated on single neuron responses (Yanai at al. 2007), we showed a robust and dynamic signal and noise correlation between functionally connected CS sites. Here, we explored CS interactions using event-related population signals. Using this approach, we were able to identify the flow of information between the 2 sites around events that tend to invoke only meager single unit responses. The observed correlations between cortical and spinal eLFPs and eMUAs were dynamic both in terms of time lag and magnitude.

First, we observed a shift in the correlation time lag of both eLFP pairs and eMUA pairs. Around Trial onset and Target onset the correlation peak time showed a mode around 0-ms lag indicating coactivation of MC and SC. For eMUAs, this coactivation was specific to MI-SC site pairs. A possible explanation for this result is a third site that (directly or indirectly) provides both MC and SC with event-related information. Several PM areas were shown (He et al. 1993; Dum and Strick 1996) to project to both MI and SC. Such areas may also affect spinal circuitry directly or indirectly via the reticulospinal tract (Kuypers and Brinkman 1970). Such a common source of activation could yield a correlation between cortical and spinal responses that straddles 0 lag. This hypothesis is not supported by our data, possibly indicating that these effects are too weak (or too fast) to be detected using population measures. Alternatively, this coactivation drive might be emitted by PM sites that were not explored in this study such as the supplementary motor area or the cingulate cortex (Dum and Strick 1996).

In contrast to the pattern of interaction observed in early stages of the task, the observed correlation shifted toward positive lags around torque onset, consistent with MC preceding SC. This may indicate that around torque onset CS interactions are dominated by a feed-forward mode of operation where MC has a paramount impact on spinal activation, thus resulting in tighter (yet delayed) locking of responses. Previous studies (Pierrot-Deseilligny and Meunier 1998; Rudomin and Schmidt 1999; Seki et al. 2003) have indeed shown that during motor action afferent inputs are suppressed, so that the relative contribution of MC to spinal circuitry is expected to increase. Here we used simultaneous recordings of MC and SC so that the functional implications of suppressed peripheral volley could be observed directly.

In addition to the change in the temporal properties of the CS correlation, we found that as the trial progressed, the magnitude of the correlation gradually increased and was maximal following torque onset. This result indicates that the responses of MC and SC are increasingly coupled during trial performance. The monotonous increase in correlation magnitude was not necessarily paralleled by an increase in the response magnitude of each site, thus suggesting that the observed effect reflects a tighter locking between cortical and spinal responses. This dynamic mode of interactions may provide the appropriate substrate for modifying ensued motor actions in response to relevant information provided to the system early in the trial.

Supplementary Material

Supplementary material can be found at: http://www.cercor.oxfordjournals.org/.

Funding

The Israel Science Foundation grants (ISF-555/01 and ISF-1355/05); The United States–Israel Binational Science Foundation (BSF-2007442); the Baruch Foundation.

Conflict of Interest: None declared.

References

Alexander
GE
Selective neuronal discharge in monkey putamen reflects intended direction of planned limb movements
Exp Brain Res
 , 
1987
, vol. 
67
 (pg. 
623
-
634
)
Alexander
GE
Crutcher
MD
Neural representations of the target (goal) of visually guided arm movements in three motor areas of the monkey
J Neurophysiol
 , 
1990
, vol. 
64
 (pg. 
164
-
178
)
Ashe
J
Georgopoulos
AP
Movement parameters and neural activity in motor cortex and area 5
Cereb Cortex
 , 
1994
, vol. 
4
 (pg. 
590
-
600
)
Bonnet
M
Requin
J
Stelmach
GE
Changes in electromyographic responses to muscle stretch, related to the programming of movement parameters
Electroencephalogr Clin Neurophysiol
 , 
1991
, vol. 
81
 (pg. 
135
-
151
)
Bortoff
GA
Strick
PL
Corticospinal terminations in two new-world primates: further evidence that corticomotoneuronal connections provide part of the neural substrate for manual dexterity
J Neurosci
 , 
1993
, vol. 
13
 (pg. 
5105
-
5118
)
Brunia
CH
Scheirs
JG
Haagh
SA
Changes of Achilles tendon reflex amplitudes during a fixed foreperiod of four seconds
Psychophysiology
 , 
1982
, vol. 
19
 (pg. 
63
-
70
)
Buchwald
JS
Halas
ES
Schramm
S
Comparison of multiple-unit and electro-encephalogram activity recorded from the same brain sites during behavioral conditioning
Nature
 , 
1965
, vol. 
205
 (pg. 
1012
-
1014
)
Bullock
TH
Signals and signs in the nervous system: the dynamic anatomy of electrical activity is probably information-rich
Proc Natl Acad Sci USA
 , 
1997
, vol. 
94
 (pg. 
1
-
6
)
Caminiti
R
Ferraina
S
Mayer
AB
Visuomotor transformations: early cortical mechanisms of reaching
Curr Opin Neurobiol
 , 
1998
, vol. 
8
 (pg. 
753
-
761
)
Churchland
MM
Afshar
A
Shenoy
KV
A central source of movement variability
Neuron
 , 
2006
, vol. 
52
 (pg. 
1085
-
1096
)
Cisek
P
A parallel framework for interactive behavior
Prog Brain Res
 , 
2007
, vol. 
165
 (pg. 
475
-
492
)
Crutcher
MD
Alexander
GE
Movement-related neuronal activity selectively coding either direction or muscle pattern in three motor areas of the monkey
J Neurophysiol
 , 
1990
, vol. 
64
 (pg. 
151
-
163
)
Donchin
O
Gribova
A
Steinberg
O
Bergman
H
Cardoso de Oliveira
S
Vaadia
E
Local field potentials related to bimanual movements in the primary and supplementary motor cortices
Exp Brain Res
 , 
2001
, vol. 
140
 (pg. 
46
-
55
)
Dum
RP
Strick
PL
Spinal cord terminations of the medial wall motor areas in macaque monkeys
J Neurosci
 , 
1996
, vol. 
16
 (pg. 
6513
-
6525
)
Fagg
AH
Arbib
MA
Modeling parietal-premotor interactions in primate control of grasping
Neural Network
 , 
1998
, vol. 
11
 (pg. 
1277
-
1303
)
Gail
A
Brinksmeyer
HJ
Eckhorn
R
Perception-related modulations of local field potential power and coherence in primary visual cortex of awake monkey during binocular rivalry
Cereb Cortex
 , 
2004
, vol. 
14
 (pg. 
300
-
313
)
He
SQ
Dum
RP
Strick
PL
Topographic organization of corticospinal projections from the frontal lobe: motor areas on the lateral surface of the hemisphere
J Neurosci
 , 
1993
, vol. 
13
 (pg. 
952
-
980
)
Kakei
S
Hoffman
DS
Strick
PL
Muscle and movement representations in the primary motor cortex
Science
 , 
1999
, vol. 
285
 (pg. 
2136
-
2139
)
Kakei
S
Hoffman
DS
Strick
PL
Direction of action is represented in the ventral premotor cortex
Nat Neurosci
 , 
2001
, vol. 
4
 (pg. 
1020
-
1025
)
Kalaska
JF
Caminiti
R
Georgopoulos
AP
Cortical mechanisms related to the direction of two-dimensional arm movements: relations in parietal area 5 and comparison with motor cortex
Exp Brain Res
 , 
1983
, vol. 
51
 (pg. 
247
-
260
)
Kalaska
JF
Scott
SH
Cisek
P
Sergio
LE
Cortical control of reaching movements
Curr Opin Neurobiol
 , 
1997
, vol. 
7
 (pg. 
849
-
859
)
Kohn
A
Smith
MA
Stimulus dependence of neuronal correlation in primary visual cortex of the macaque
J Neurosci
 , 
2005
, vol. 
25
 (pg. 
3661
-
3673
)
Komiyama
T
Tanaka
R
The differences in human spinal motoneuron excitability during the foreperiod of a motor task
Exp Brain Res
 , 
1990
, vol. 
79
 (pg. 
357
-
364
)
Kuypers
HG
Brinkman
J
Precentral projections to different parts of the spinal intermediate zone in the rhesus monkey
Brain Res
 , 
1970
, vol. 
24
 (pg. 
29
-
48
)
Lebedev
MA
Wise
SP
Insights into seeing and grasping: distinguishing the neural correlates of perception and action
Behav Cogn Neurosci Rev
 , 
2002
, vol. 
1
 (pg. 
108
-
129
)
Luppino
G
Rizzolatti
G
The organization of the frontal motor cortex
News Physiol Sci
 , 
2000
, vol. 
15
 (pg. 
219
-
224
)
Minelli
A
Marzi
CA
Girelli
M
Lateralized readiness potential elicited by undetected visual stimuli
Exp Brain Res
 , 
2007
, vol. 
179
 (pg. 
683
-
690
)
Mitzdorf
U
Current source-density method and application in cat cerebral cortex: investigation of evoked potentials and EEG phenomena
Physiol Rev
 , 
1985
, vol. 
65
 (pg. 
37
-
100
)
Moll
L
Kuypers
HG
Premotor cortical ablations in monkeys: contralateral changes in visually guided reaching behavior
Science
 , 
1977
, vol. 
198
 (pg. 
317
-
319
)
Mushiake
H
Tanatsugu
Y
Tanji
J
Neuronal activity in the ventral part of premotor cortex during target-reach movement is modulated by direction of gaze
J Neurophysiol
 , 
1997
, vol. 
78
 (pg. 
567
-
571
)
O'Leary
JG
Hatsopoulos
NG
Early visuomotor representations revealed from evoked local field potentials in motor and premotor cortical areas
J Neurophysiol
 , 
2006
, vol. 
96
 (pg. 
1492
-
1506
)
Pierrot-Deseilligny
E
Meunier
S
Rudomin
P
Romo
R
Mendell
L
Differential control of presynaptic inhibition of Ia terminals during voluntary movement in humans
Presynaptic inhibition and neural control
 , 
1998
New York
Oxford University Press
(pg. 
351
-
365
)
Prut
Y
Fetz
EE
Primate spinal interneurons show pre-movement instructed delay activity
Nature
 , 
1999
, vol. 
401
 (pg. 
590
-
594
)
Riehle
A
Requin
J
The predictive value for performance speed of preparatory changes in neuronal activity of the monkey motor and premotor cortex
Behav Brain Res
 , 
1993
, vol. 
53
 (pg. 
35
-
49
)
Roux
S
MacKay
WA
Riehle
A
The pre-movement component of motor cortical local field potentials reflects the level of expectancy
Behav Brain Res
 , 
2006
, vol. 
169
 (pg. 
335
-
351
)
Rubino
D
Robbins
KA
Hatsopoulos
NG
Propagating waves mediate information transfer in the motor cortex
Nat Neurosci
 , 
2006
, vol. 
9
 (pg. 
1549
-
1557
)
Rudomin
P
Schmidt
RF
Presynaptic inhibition in the vertebrate spinal cord revisited
Exp Brain Res
 , 
1999
, vol. 
129
 (pg. 
1
-
37
)
Sawaguchi
T
Yamane
I
Kubota
K
Application of the GABA antagonist bicuculline to the premotor cortex reduces the ability to withhold reaching movements by well-trained monkeys in visually guided reaching task
J Neurophysiol
 , 
1996
, vol. 
75
 (pg. 
2150
-
2156
)
Schoffelen
JM
Oostenveld
R
Fries
P
Neuronal coherence as a mechanism of effective corticospinal interaction
Science
 , 
2005
, vol. 
308
 (pg. 
111
-
113
)
Seki
K
Perlmutter
SI
Fetz
EE
Sensory input to primate spinal cord is presynaptically inhibited during voluntary movement
Nat Neurosci
 , 
2003
, vol. 
6
 (pg. 
1309
-
1316
)
Shen
L
Alexander
GE
Neural correlates of a spatial sensory-to-motor transformation in primary motor cortex
J Neurophysiol
 , 
1997
, vol. 
77
 (pg. 
1171
-
1194
)
Tanji
J
Kurata
K
Contrasting neuronal activity in supplementary and precentral motor cortex of monkeys. I. Responses to instructions determining motor responses to forthcoming signals of different modalities
J Neurophysiol
 , 
1985
, vol. 
53
 (pg. 
129
-
141
)
Wise
SP
The primate premotor cortex fifty years after Fulton
Behav Brain Res
 , 
1985
, vol. 
18
 (pg. 
79
-
88
)
Yanai
Y
Adamit
N
Harel
R
Israel
Z
Prut
Y
Connected corticospinal sites show enhanced tuning similarity at the onset of voluntary action
J Neurosci
 , 
2007
, vol. 
27
 (pg. 
12349
-
12357
)
Yanai
Y
Adamit
N
Israel
Z
Harel
R
Prut
Y
Coordinate transformation is first completed downstream of primary motor cortex
J Neurosci
 , 
2008
, vol. 
28
 (pg. 
1728
-
1732
)

Author notes

Itay Asher and Nofya Zinger contributed equally to the study.