INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Many motor control models assume that making a voluntary motor response to a sensory signal involves a number of central information-processing stages that culminate in the activation of muscles (Flanders et al. 1992; Ghez et al. 1997; Gordon et al. 1994; Kalaska and Crammond 1992; Requin et al. 1988, 1993; Riehle et al. 1994; Rosenbaum 1983). They also generally assume that at least some of these sensorimotor processes are serially ordered and often arbitrarily group them into two broad sequential stages of planning and execution. Neurophysiological studies in the cerebral cortex have shown that, while this sequential model is simplistic, it still retains some heuristic value for describing the general information-processing structure of cortical motor control mechanisms (Andersen et al. 1997; Bastian et al. 1998; Crammond and Kalaska 1994; Fu et al. 1993, 1995; Georgopoulos et al. 1986; Graziano and Gross 1998; Kalaska et al. 1997; Kurata 1993; Pellizzer et al. 1995; Requin et al. 1988, 1993; Riehle 1991; Riehle and Requin 1989; Riehle et al. 1994; Schall and Bichot 1998; Shen and Alexander 1997a,b; Wise et al. 1996-1998; Zhang et al. 1997).

In multiple-choice reaction-time (RT) tasks, for instance, these processing stages are presumably initiated or accomplished during the behavioral reaction time between the appearance of the signal that defines the appropriate response and the onset of muscle activity. However, it is widely assumed that early planning stages involve cognitive processses, such as identification of stimulus properties and saliency, and response selection, that are initiated by the sensory signal but are not causally related to muscle activation and so are theoretically dissociable from overt motor output (Lecas et al. 1986; Rosenbaum 1983). This provides the rationale for instructed-delay (ID) tasks in which a CUE signal provides information about a desired movement whose execution must be delayed until a subsequent GO signal. This should allow the subject to plan the signaled attributes of the ensuing movement and consequently reduce the amount of information processing needed after the GO signal.

To the extent that this sequential framework is valid, it leads to two simple predictions about the effect of prior information on neural events in RT and ID tasks. First, neuronal correlates of some early planning stages that occur after the GO signal in RT tasks should instead be evoked by the CUE signal in ID tasks. At the single-cell level, this could take the form of discharge with similar task-related properties, such as similar directional tuning in a multi-directional task. Second, these early planning events need not be recapitulated after the GO signal in ID trials. This should result in a modification of the post-GO cell discharge in the ID task compared with that recorded in the RT task.

Although these assumptions about prior information and neuronal activity are widely held, they have not been put to a systematic test. To do so, one must compare the responses of cells between RT and ID tasks. However, there have been few studies of precentral cortex activity under both conditions, and they have focused mainly on properties of delay period discharge (Georgopoulos et al. 1989; Smyrnis et al. 1992). Analysis of post-GO activity was less detailed, and no major effects of prior information were reported (Smyrnis et al. 1992). Another study did report significant reductions in post-GO activity of some cells as a function of prior information provided in ID trials (Riehle and Requin 1989). Although all trials in that study were ID trials, the instructional signal in one trial class was completely ambiguous, so that the monkeys could not plan any attributes of the intended response before the GO signal. Riehle and Requin (1989) also surveyed a much larger area of the precentral gyrus than did Smyrnis et al. (1992). Finally, whereas Riehle and Requin (1989) provided varying degrees of prior information for two opposite directions of isolated wrist movements, the other studies always provided complete prior information for reaching movements of the whole limb in eight different directions (Georgopoulos et al. 1989; Smyrnis et al. 1992). Therefore differing levels of prior information, response uncertainty, and response complexity may have all contributed to the seemingly conflicting results.

The present study examines these widespread assumptions of cognitive motor control models. We recorded from arm-related cells in the primary motor (MI) and dorsal premotor (PMd) cortex of monkeys during visually guided reaching movements in both RT and ID tasks. In RT trials, no information about response metrics was provided until the appearance of the GO signal. In contrast, the CUE signal in ID trials completely specified the intended movement, including both direction and target location. Unlike RT trials, the GO signal in ID trials served only as a timing signal that provided no additional response-related information beyond that furnished by the CUE. In theory, the monkeys could completely plan the movement during the IDP of ID trials and generate only execution-related cell activity after the GO signal.

The results were consistent with both predictions. Nevertheless, they also showed that the serially ordered dichotomy between movement planning and execution, that can be absolute from a control theoretical perspective, is confounded in the discharge of single precentral neurons (Requin et al. 1988, 1993; Shen and Alexander 1997a,b; Zhang et al. 1997). Some features of the post-GO responses of these same MI and PMd cells during RT trials have already been described in detail (Crammond and Kalaska 1996), as have some aspects of their activity during ID trials (Crammond and Kalaska 1989a, 1994; Kalaska and Crammond 1990, 1995).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

The subjects were one female and two male macaque monkeys (Macaca mulatta) weighing 4.5, 5.0, and 5.5 kg, respectively. All procedures and animal care respected Canadian Medical Research Council guidelines and the National Institutes of Health Guide for the Care and Use of Laboratory Animals.

Experimental design

The task apparatus consisted of a horizontal target panel over which was suspended a pendulum with a handle at its free end (Kalaska et al. 1989). The target panel contained nine triplets of miniature red, green, and yellow light-emitting diodes (LEDs), with one triplet at the center and eight triplets distributed evenly around it in a circle of 8-cm radius (Fig. 1). Only the red and green LEDs were used in the tasks described here. The monkeys were trained to hold the handle over whichever red LED was illuminated. Handle position was measured ultrasonically every 10 ms (Graf/Pen 3, Science Accessories).

View larger version (23K): [in this window] [in a new window]   Fig. 1. Cartoons of the target panel illustrating the sequence of stimulus events in reaction-time (RT) trials and 3 classes of instructed-delay trials: direct-delay (DD), direct-delay with a non-spatial GO signal (DD-NS), and memorized-delay (MEM). Small dots, target locations; solid circle, red light-emitting diode (LED); open circle, green LED; square box, desired position of the handle. Trial epochs denoted as Pre-CUE: epoch before the onset of the CUE stimulus (left vertical dashed line) in instructed delay trials; Instructed Delay period: from the appearance of the CUE to the appearance of the GO signal (right vertical dashed line); Reach and Target hold: from the appearance of the GO signal to the end of the trial. In DD and DD-NS trials, the instruction CUE remained illuminated for the duration of the IDP. In MEM trials, the instruction CUE was illuminated for only the 1st 500 ms of the IDP. In RT trials, "CUE on" corresponded to the time an event marker was entered in the trial record although no CUE was presented.

The principal task involved two classes of trials: RT trials and a type of ID trial [direct-delay (DD) trials; Fig. 1]. Two monkeys also performed other types of ID trials described here and elsewhere (Crammond and Kalaska 1994; Kalaska and Crammond 1995). All trials began when the central red LED was illuminated. The monkey held the handle over this LED for a variable time period, until it was extinguished and one of the peripheral red target LEDs was illuminated (GO signal). The monkey moved the handle to the red target LED and held it there for a fixed period of 2 s. In RT trials, no other signals appeared, so that the monkey did not know the direction of movement until the GO signal appeared (Fig. 1). In DD trials, in contrast, a green LED (the instruction CUE signal) was illuminated at one of the eight peripheral locations while the monkey was holding the handle at the center (Fig. 1). The green CUE indicated the direction of the impending movement and remained on until the GO signal, when it and the central LED were extinguished and the red LED at the CUE location was illuminated.

The two trial classes and eight movement directions were presented in a randomized-block sequence (Snedecor and Cochran 1980). As a result, neither movement direction nor trial class were predictable when a trial began. The only exception occurred after trials in which the monkeys made an error, when a trial of the same class and direction was repeated. Three complete replications of all combinations of direction and class were required for statistical testing. Most data files contained 5-10 replications.

Trial epochs

Each trial was divided into three sequential parts: the Center-hold time (CHT) before the GO signal, the Reach epoch from the GO signal to the end of movement, and the Target-hold time (THT; Fig. 1). The CHT of ID trials was further divided into pre-CUE and instructed-delay period (IDP) epochs, before and after the presentation of the CUE signal. The duration of each varied randomly between 1 and 3 s. We defined two further components of the IDP epoch in DD trials: the early-IDP epoch (the 1st 500 ms after CUE presentation) and the late-IDP epoch (the last 500 ms immediately preceding the GO signal). The CHT of RT trials likewise comprised arbitrary "pre-CUE" and "IDP" epochs of 1-3 s duration, with a time marker inserted into the trial record when a CUE would have been presented had it been a DD trial. The interval between the time marker and the GO signal in RT trials will be called the noninstructed delay period (NIDP) epoch. This ensured the same range of CHT durations in RT and DD trials. It also permitted comparison of activity at equivalent times from the beginning of the trial, to distinguish nonspecific time-dependent changes in discharge during the CHT (Crammond and Kalaska 1996; Vaadia et al. 1988) from responses elicited by the instructional CUEs.

The Reach epoch was further divided into reaction-time (RTE) and movement-time (MTE) epochs. A movement-velocity curve was generated for each trial by differentiation of the handle's X-Y coordinates measured every 10 ms. A recursive algorithm (Kalaska et al. 1989) used the velocity curve to detect movement onset, and its end when the handle became stationary over the target.

Task variants

The red GO signal was identical in RT and DD trials, to exclude the possibility that task-dependent changes in post-GO activity between trial classes could be attributed to physical differences in the stimuli. As a result, however, the monkeys could conceivably ignore the CUE in DD trials, plan no properties of the movement during the IDP, and only use the target information provided by the GO signal. To address this unlikely possibility, two modified DD trial types were introduced that could only be performed correctly if the monkey used the prior information provided by the CUE. In the first variant, monkeys performed DD trials in which a nonspatial GO signal (DD-NS) was provided by the simultaneous illumination of all eight peripheral red LEDs (Fig. 1). DD-NS trials were presented in separate blocks of trials interleaved with standard DD trials. In the second variant, DD-NS trials were interleaved with memorized-delay (MEM) trials (Fig. 1). In MEM trials, the CUE was illuminated for only the first 500 ms of the IDP epoch. Only nonspatial GO signals were used in MEM trials. Thus in DD-NS trials, monkeys had to use the CUE to identify the correct target. In MEM trials, there was a further requirement that the CUE location had to be memorized. These task variants were tested on selected cells in two monkeys.

Behavioral control

The monkeys' behavior was closely monitored to minimize overt anticipatory movements before the GO signal, especially during the IDP epoch of DD trials. The monkeys held the handle within a small window of 4- to 5-mm radius over the central LED during the CHT. The handle's position was displayed on a monitor at ×10 magnification. Any noticeable drift of the handle toward a target before the GO signal, even if remaining within the central window, resulted in manual interruption of the trial. Furthermore, the monkeys were always observed by one of the experimenters during data collection, and a trial was halted if a change in arm or trunk posture was detected during the CHT, even if this did not cause the handle to leave the central window. After the GO signal, the monkey had to exit the central window in no less than 150 ms to eliminate anticipatory movements before the GO signal, and no more than 600-750 ms to ensure attention to the task and prompt responses. The monkeys were trained to a success rate of 70-90%. Thereafter, most errors were due to the difficulty in holding the handle within the small central window for the long duration of the CHT. Overt anticipatory movements during data collection were rare but readily detected and extinguished by the methods just described.

EMG recordings

Stable performance in the tasks was further evaluated by recording the task-related EMG activity of all the major muscles of the shoulder joint and shoulder girdle as well as several axial, paraspinal, and neck muscles. This was undertaken at various times prior to, during, and after the several months of neuronal data collection in each monkey. Pairs of fine, 40-gauge Teflon-insulated stainless steel wires were inserted percutaneously into the bellies of selected muscles using 30-gauge hypodermic needles. All electromyographic (EMG) activity was amplified (×1,000 to ×5,000), filtered (100 Hz to 3 kHz), rectified, and integrated (10 ms bin duration) before storage. The identity of the implanted muscles was verified by observation of EMG activity outside of the task and by microstimulation of the implanted muscles via the recording electrodes. If microstimulation failed to evoke a palpable local contraction of the desired muscle belly or the expected joint motions, the electrodes were removed and re-inserted.

Neuronal data collection

After training, the monkeys were surgically prepared for data collection. Using standard aseptic techniques and barbiturate anesthesia (35 mg/kg iv), a trephine hole was opened in the skull over the precentral gyrus contralateral to the performing arm. A Plexiglas recording chamber was fixed over the craniotomy using vitallium screws and neurosurgical acrylic cement. The chamber was positioned to span the precentral cortex between the central and arcuate sulci. A stainless steel head-fixation post was also embedded in the acrylic.

Daily recording sessions began after a postoperative recovery period of 10 days during which prophylactic antibiotics and analgesic drugs were administered. Standard chronic extracellular recordings were made using glass-insulated platinum-iridium electrodes (Crammond and Kalaska 1996; Kalaska et al. 1989). The discriminated, extracellular spike activity of single neurons was recorded during performance of the task and also tested for response properties outside of the task. To be included in the database, the activity of a cell had to meet two criteria. First, neuronal discharge had to be related to the proximal arm or shoulder girdle on the basis of responses to passive manipulations and to spontaneous active movements of forelimb segments. Cells judged to be related to the distal arm or trunk were not studied further. Second, cell activity had to change during one or more epochs of the RT or DD trials whether or not that modulated activity appeared to be directionally tuned. At the end of certain penetrations, microlesions (10 µA, 10-20 s) were made in the cortex at specific locations along the electrode track. At the end of each daily recording session, the cylinder was cleaned, flushed with sterile saline, and closed.

Data collection lasted 8-10 wk in each chamber. When the experiments were completed, the monkeys were deeply anesthetized and perfused with saline and then 10% Formalin solutions. The dura was removed, and dissecting pins were inserted in the brain at known coordinates to delimit the cortical region studied. Using the pins as cutting guides, the cortex was blocked and 30 µm frozen sections were cut, stained with cresyl violet, and examined by light microscopy to locate the microelectrode penetrations.

Data analysis

A three-way ANOVA (task, direction, and replications; P < 0.01) (Snedecor and Cochran 1980) was applied to the combined data from both tasks for a given cell or EMG record. However, interpretation of the three-way ANOVA results was complex. In particular, different cells with clear directionally tuned IDP epoch activity in DD trials showed all possible combinations of significant main effects of task and direction, and task-direction interactions during the IDP.

Therefore to identify directionally tuned responses during the various trial epochs in each task, we used two statistical tests on the data from RT and DD tasks separately. A two-way ANOVA (direction and replications, P < 0.01) was used to detect statistically different responses at one or more directions of movement independent of their directional pattern. A second, nonparametric "bootstrapping" test was used to identify a significant unimodal bias in cell activity (Crammond and Kalaska 1996; Lurito et al. 1991). First, the degree of directional bias in a cell's task-related activity was determined by calculating the mean length of the variation of its discharge across all eight movement directions (Fortier et al. 1989; Lurito et al. 1991; Mardia 1972). The mean length of a cell that discharged uniquely for 1 movement direction would be 1.0, whereas that of a cell with uniform activity across all 8 directions would be 0.0. Bootstrapping was then used to assess whether that directionally tuned activity could have arisen by chance. The set of trials was randomly reassigned to different "movement directions," and the mean length of the directional tuning of the reshuffled trials was calculated. This bootstrap reshuffling procedure was repeated up to 4,000 times to calculate 4,000 randomly selected mean lengths, which were compared with the mean length of the cell's task-related response pattern. If fewer than 40 reshuffled-data mean lengths exceeded the mean length of the cell's task-related response, the cell was classified as directionally tuned (approximate P < 0.01) (Lurito et al. 1991). A response in any epoch was scored as directional only if both tests were significant.

Even at the 1% level, the ANOVA and bootstrap tests were very sensitive, and changes in cell activity that looked weak on visual inspection could prove to be significant. Therefore as a separate quantitative measure of the degree of directional tuning in each epoch, a directional dynamic range of activity was calculated, defined as the difference between the maximum and minimum averaged discharge rates recorded at different directions in a given epoch.

A cell's preferred direction was calculated in each epoch using trigonometric moments (Mardia 1972). Since cell activity was never perfectly uniform in all eight directions, one can calculate a "preferred direction" even for activity that is not statistically directional. In such cases, the preferred direction is the weighted center of the random fluctuations in discharge.

A sliding-window procedure was used to compare the post-GO responses of single cells between RT and DD trials. Data in all trials at a cell's preferred direction were aligned to the onset of the GO signal. The mean discharge rate of the cell, including partial spike intervals, was calculated within a 100-ms time window that was stepped forward in 10-ms intervals. The maximum mean 100-ms discharge rate was determined in two different time periods. The first was from 100 to 300 ms post-GO, corresponding approximately to the period of task-related discharge during the RTE of RT trials (Crammond and Kalaska 1996). The second period was from 300 to 1,000 ms post-GO, which includes the MTE and the first part of the THT. This analysis was performed separately for the RT and DD trials, using the preferred direction of cell activity in RT trials.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

EMG recordings

Fifty-seven sets of EMG activity were recorded from the major proximal-arm muscles [deltoids (3 heads), pectoralis major, latissimus dorsi, triceps longus, teres major, supraspinatus, infraspinatus, and subscapularis], and 19 sets were taken from axial muscles (rostral and caudal trapezius, thoracic and cervical paraspinals, rhomboids, splenius capitis, and atlantoscapularis anterior). No directionally tuned changes in activity were measured during the IDP of DD trials in 71/76 EMG data sets [Table 1; see Kalaska and Crammond (1995) for examples of EMG records]. Two data sets exhibited a weak directional change in the late IDP of DD trials (  +, Table 1), and two others exhibited directional EMG activity only when averaged over the entire IDP (+  , Table 1). In all four cases, EMG records from the same muscles in the same animals at other times were not significant, suggesting that they did not reflect a systematic anticipatory strategy. The fifth significant result came from one monkey that made idiosyncratic exaggerated licking and sucking movements on the juice reward tube after the appearance of any CUE. This evoked strong cyclic bursts of activity in the splenius capitis muscle at all times in the IDP of DD trials (+ + +, Table 1). No corresponding rhythmic activity was recorded in any other muscles or any neurons studied in that monkey.

Neuronal data set

We recorded the task-related activity from 503 neurons, including 279 in the PMd and 224 in MI. The border between PMd and MI was placed at that point rostral to which standard intracortical microstimulation (ICMS) failed to evoke visible movements or muscle contractions at currents up to 50 µA (Crammond and Kalaska 1996). Histology confirmed that the cortex rostral to this point contained very few large pyramidal cells (>29 µm) in all three monkeys. MI was further subdivided into rostral (MIr, 72 cells) and caudal (MIc, 152 cells) zones. MIc was the part of area 4 forming the anterior bank of the central sulcus, in which ICMS currents as low as 3-5 µA would evoke brisk localized muscle contractions. MIr occupied the cortex from the lip of the central sulcus to the PMd border near the superior precentral sulcus. In MIr, ICMS evoked muscle twitches at fewer locations than in MIc, and thresholds were rarely less than 10-15 µA. However, there were no abrupt transitions in ICMS thresholds across the precentral gyrus, so that the locations of the borders between the three areas were somewhat arbitrary. Maps of penetration sites and borders between areas can be found elsewhere (Crammond and Kalaska 1996).

Temporal response patterns at the preferred direction during the IDP

The most common response profiles in PMd during the IDP of DD trials were sustained tonic activity changes, or incrementing or decrementing ramp changes in discharge (Fig. 3, B and D). The IDP activity of 135 PMd cells was principally of this type. A second response type seen in 44 PMd cells was a short-latency phasic response after the appearance of the CUE signals, with little activity for the rest of the IDP (Fig. 2D). A further 48 PMd cells showed combinations of phasic and sustained responses (Fig. 2B). The remaining 52 cells showed no IDP activity or were unclassifiable. The phasic and sustained responses have been called "signal" and "set" responses (Weinrich and Wise 1982; Weinrich et al. 1984), but these labels were not adopted here due to their strong functional implications. Early phasic IDP responses were progressively less frequent and prominent with increasingly caudal recording sites in MIr and MIc. Finally, many cells were only active after the GO signal in both RT and DD trials. Such movement-only cells were much more common in MI than in PMd.

View larger version (36K): [in this window] [in a new window]   Fig. 2. A and B: raster plots illustrating the discharge of a single dorsal premotor cortex (PMd) cell during RT (A) and DD (B) trials in 8 different directions, as indicated by the cartoons of the target panel between the 2 sets of rasters. All trials have been aligned to the time of appearance of the CUE signal in DD trials or to the corresponding time marker in RT trials (left vertical dotted line and arrows), and to the GO signal (right vertical dotted line and arrows). For each raster line, the 2 thick markers to the right of the GO signal indicate the time of onset and end of movement for that particular trial. Horizontal calibration bar, 1,000 ms. C and D: raster representation of the discharge of a different PMd cell. Same format as in A and B.

Directional tuning of MI and PMd cell activity in RT and DD trials

Figures 2 and 3 illustrate the range of directional responses of PMd cells in RT and DD trials. The cells in Fig. 2 showed fairly consistent directional tuning across different epochs in RT and DD trials, while those in Fig. 3 were more complex. The first neuron emitted a broadly tuned phasic burst of activity confined to the RTE of RT trials (Fig. 2A; preferred direction 171°). During the IDP of DD trials (Fig. 2B), the cell emitted a short-latency burst after the appearance of the CUE followed by later ramplike increases in discharge (preferred direction: early IDP, 149°; late IDP, 161°; entire IDP, 157°). The cell was significantly directionally tuned in all three IDP epochs (i.e., + + + in Table 1). The ability of this cell to generate sustained discharge during the IDP, even though it only emitted a phasic burst of activity in RT trials, was a relatively common observation in this sample of PMd cells.

View larger version (67K): [in this window] [in a new window]   Fig. 3. A raster representation of the discharge of 2 different PMd cells (A and B; C and D) in RT and DD trials. Same format as in Fig. 2.

The second cell (Fig. 2, C and D) also emitted a brief phasic burst after the GO signal in RT trials (Fig. 2C; preferred direction: RTE, 33°). However, it only emitted a phasic burst during the IDP in DD trials (Fig. 2D, preferred direction: early IDP, 55°) and was directionally tuned when averaged over the entire IDP and during the early IDP, but not during the late IDP (i.e., + +  in Table 1). The intensity of the post-GO phasic response was reduced in DD trials (Fig. 2D) compared with RT trials (Fig. 2C), unlike the cell in Fig. 2, A and B. Figure 2 illustrates how PMd cells with similar responses in RT trials could respond very differently during DD trials.

Figure 3A shows one of the 32 PMd cells whose directional tuning before (RTE) and after movement (THT) were nearly opposite (preferred direction, 147 and 310°, respectively). In DD trials, this cell was directionally tuned throughout the IDP (Table 1, + + +), with relatively stable directionality (preferred direction: entire IDP, 154°; early IDP, 125°; late IDP, 164°). This suggests that the IDP response of that cell predicted the directionality of only its earliest post-GO discharge in RT trials but not the subsequent change in directionality in later epochs.

Figure 3, C and D, shows a PMd cell with no significant directional tuning during the RT trials, but with strong reciprocal tuning throughout the IDP (+ + +, preferred direction: entire IDP, 6°; early IDP, 357°; late IDP, 3°). Overall, 45 PMd cells were nondirectional during RT trials but directional during the IDP. Such cells were very rare in MIr and MIc. Interestingly, cells with the opposite pattern, active in RT trials but inactive in DD trials, were never seen. A cell could be much less active after the GO signal in DD trials compared with RT trials, but this was invariably coupled with activity changes during the IDP (Fig. 2, C and D).

Rostrocaudal gradient of incidence of directionally tuned IDP responses

Most MIc cells showed no significant and directionally tuned activity changes during the IDP of DD trials ("movement-only" cells; Table 1,   , 71.1%). The incidence of such cells declined to 41.7% in MIr, and 14.3% in PMd. However, 11 of these 40 PMd cells showed significant activity changes that were nondirectional during the IDP, so that only 29 PMd cells (10.4%) were classed as true movement-only cells. There was a reciprocal progressive rostrocaudal decline in the incidence of cells that showed directionally tuned IDP activity in DD trials from PMd (85.7%) to MIr (58.3%) and MIc (28.9%; Table 1). Both gradients were statistically significant (P < 0.01, ² test). A three-way ANOVA showed corresponding gradients in the frequency of main effects of direction or task, and direction/task interactions during the IDP (Table 2).

The rostrocaudal decline of directionally tuned activity was greater during early IDP than late IDP. The incidence of directionally tuned cells in the early IDP (Table 1, + + +, + + ,  + +, + ) declined sharply from PMd (59.1%), to MIr (23.6%), and MIc (2.6%) (P < 0.01, ² test). During the late IDP (Table 1, + + +, +  +,  + +,  +), the decline was only from 67.4% in PMd, to 45.8% in MIr and 23.7% in MIc (P < 0.01, ² test). Alternatively, all three areas showed an increase in the incidence of directional activity in late IDP compared with early IDP, but the degree of increase was greater in MIc (from 2.6 to 23.7%) than in MIr and PMd. Finally, there was no significant difference (P > 0.05, ² test) in the incidence of cells that became directionally active only in the late IDP and not earlier (Table 1, +  +,   +) in PMd (24.0%), MIr (31.9%), and MIc (21.7%). However, those cells represented a significantly greater proportion of the cells that showed any IDP activity as one progressed caudally from PMd (67/239 cells, 28.0%), to MIr (23/42, 54.8%), and MIc (33/44, 75.0%; P < 0.01, ² test).

In summary, there was a strong rostrocaudal decline in the number of cells with directional IDP activity across the precentral gyrus (Johnson et al. 1996; Riehle and Requin 1989; Weinrich et al. 1984). There was also a strong trend for activation early in the IDP in PMd and gradually later recruitment of cells in more caudal parts of the precentral gyrus.

Rostrocaudal gradient of magnitude of directionally tuned IDP responses

There was a pronounced rostrocaudal decline in the intensity of IDP responses (Fig. 4). When the activity of each cell in RT (Fig. 4, A-C) and DD trials (Fig. 4, D-F) was oriented to its preferred direction in RT trials, a directional signal can be seen during the IDP of DD trials (Fig. 4, D-F) whose intensity declined systematically from PMd to MIc.

View larger version (32K): [in this window] [in a new window]   Fig. 4. A-F: population histograms of the summed response of all cells recorded in each cortical area, calculated at the preferred direction of each cell during the reaction-time epoch (RTE) + movement-time epoch (MTE) of RT trials () and at the opposite direction (). Population histograms are shown separately for cell activity recorded during RT (A-C) and DD trials (D-F) for each of 3 cortical regions: caudal primary motor cortex (MIc; A and D; 152 cells), rostral primary motor cortex (MIr; B and E; 72 cells), and PMd (C and F; 279 cells). For each histogram, each contributing trial was aligned both to the time of appearance of the CUE signal (DD trials) or CUE time marker (RT trials) and to the GO signal. The 2 T-bars below each histogram to the right of the GO signal alignment indicate the mean time (±SD) of onset (left T) and end (right T) of movement. Horizontal calibration bar, 500 ms; vertical calibration bar, 10 spikes/s.

This gradient was confirmed by calculating the dynamic range of single-cell IDP responses (Table 3). The dynamic range in the NIDP of RT trials is a measure of the inherent variability of cell activity when no directional information was provided. This was similar across the precentral gyrus at all times during the NIDP (Table 3A).

In contrast, the size of the dynamic range over the entire IDP of DD trials decreased progressively along the rostrocaudal axis (Table 3B). Furthermore, there was a rostrocaudal delay in the epoch at which an increase in the dynamic range was found during DD trials. In PMd, there was a large dynamic range during the early IDP of DD trials (P < 0.01, paired t-test with data from early NIDP of RT trials), that increased further in the late IDP. In MIr, the increase in the early-IDP dynamic range was modest but significant, and more pronounced in the late IDP. In MIc, a significant increase in dynamic range was only evident in the late IDP of DD trials, and not during the early IDP.

The preceding analysis included the cells without significant directionally tuned IDP activity. The analysis was repeated for only the cells with significant IDP activity. This had a relatively minor effect on the PMd data, but the IDP responses appeared noticeably larger in MIr and especially in MIc (Table 3C; Fig. 5). Nevertheless, the mean dynamic range during the entire IDP still declined from PMd to MIr and MIc. As for the timing gradient, the rostrocaudal decrease in dynamic range was still pronounced in the early IDP, especially the progressive reduction of the initial short-latency phasic response after the CUE from PMd to MIc (Fig. 5, A-F). In the late IDP, the difference in the mean dynamic range was not as pronounced between areas for those cells with significant IDP responses (Table 3C) as for the entire population (Table 3B). Thus by the end of the IDP, the intensity of significant single-cell IDP responses was nearly equal across the precentral gyrus, but far fewer cells generated those responses in MIc than in PMd. Furthermore, directional IDP responses began abruptly after the CUE in PMd (Fig. 5, C and F) but evolved much more gradually during the delay period in MIc (Fig. 5, A and D).

View larger version (34K): [in this window] [in a new window]   Fig. 5. A-C: population histograms of the summed response of all cells recorded in each cortical area, calculated at the preferred direction of each cell during the instructed-delay period (IDP) epoch of DD trials () and at the opposite direction (). D-F: population histograms formatted as in A-C but now only including the activity of those cells recorded in each of MIc (D; 44 cells), MIr (E; 42 cells), and PMd (F; 239 cells) that had significant directionally tuned activity at any time during the IDP epoch of DD trials (1st 7 columns of Table 1).

Comparison of directional tuning properties of cell activity in RT and DD trials

An important objective of this study was to compare the directional tuning of post-GO activity in RT tasks with that during the IDP and post-GO epochs of DD trials.

The directionality of post-GO movement-related activity was very similar during the Reach epoch of RT and DD trials (Fig. 6; mean arithmetic angular differences). The directional correlation nevertheless declined systematically from MIc to PMd (Fig. 6; mean absolute angular differences and correlations). Interestingly, for the 29 movement-only cells in PMd, the post-GO directionality in RT and DD trials was as highly correlated (mean absolute angular difference 11.6°, r = 0.95) as in MIc, whereas the post-GO directional correlation was lower (mean angular difference 31.4°, r = 0.73) for the remaining PMd cells that had significant IDP responses. Part of this increased scatter was attributable to the 45 PMd cells that were not significantly tuned in RT trials and whose tuned IDP activity in DD trials carried over into the RTE (Fig. 3D).

View larger version (22K): [in this window] [in a new window]   Fig. 6. Scatter plots illustrating the correlation between the directional tuning of cell activity recorded during the combined RTE + MTE epoch of RT trials (abcissa) and of DD trials (ordinate) for all cells recorded in MIc (A; 152 cells), MIr (B; 72 cells), and PMd (C; 279 cells). Each symbol represents the paired data from a single cell. Diagonal dashed line, the identity line. Values below each graph indicate the arithmetic (signed) and absolute (unsigned) mean angular difference for each data set and their correlation.

The directional tuning of cell activity was next compared between the IDP of DD trials and the post-GO Reach epoch of RT trials (Fig. 7). There was greater directional variability between those epochs than there was between the Reach epochs of DD and RT trials (Fig. 6). Nevertheless, the tuning between those two epochs was correlated for the whole sample in PMd (Fig. 7C) and in MIr (Fig. 7B). However, in MIc the correlation was poor (Fig. 7A). The correlation improved in MIc and MIr when only those cells with significant directionally tuned IDP activity were considered (squares in Fig. 7, A-C). Nevertheless, the correlation with directionality in the Reach epoch was still weaker for MIc than MIr and PMd.

View larger version (28K): [in this window] [in a new window]   Fig. 7. Scatter plots illustrating the correlation between the directional tuning of cell activity recorded during the combined RTE + MTE epoch of RT trials (abcissa) and the IDP epoch of DD trials (ordinate) for all cells recorded in MIc (A; 152 cells), MIr (B; 72 cells), and PMd (C; 279 cells). , cells with significant directionally tuned activity recorded during the IDP epoch of DD trials (1st 7 columns of Table 1). +, cells with nonsignificant directional tuning during the IDP epoch of DD trials (last column of Table 1). Values below each graph indicate the absolute mean angular difference for each data set and their correlation, for the total sample and for only those cells with directionally tuned IDP activity.

The preferred direction of significantly tuned IDP activity in DD trials was next compared with the preferred direction of activity for each of the three successive post-GO behavioral epochs in RT trials (Table 4). In both PMd and MIr, the correlation was highest between the IDP of DD trials and the RTE of RT trials prior to movement onset, followed by an abrupt decline after movement onset (Table 4). In contrast, the correlation with RTE tuning was the weakest in MIc, but remained fairly stable in the subsequent MTE and THT epochs.

The population histograms of Figs. 4 and 5 illustrate the differences in directionality between the IDP and the post-GO period in a different manner. In Fig. 4, D-F, cell activity recorded in DD trials is oriented to the preferred direction of cell discharge during the Reach epoch of RT trials. Because the movement-period directionality was highly correlated between trial classes (Fig. 6), strong and approximately reciprocally tuned activity can be seen for opposite movement directions after the GO signal in DD trials in all three areas. A directionally tuned signal was also clear during the IDP of DD trials in PMd (Fig. 4F), but less so in MIr (Fig. 4E) and appeared very modest in MIc (Fig. 4D). When DD trial data were re-oriented to the preferred direction of activity during the IDP of DD trials (Fig. 5), a directionally tuned IDP response became more pronounced in all three areas, but especially in MIc. However, whereas the post-GO histograms in PMd were relatively unaffected by this reorientation (compare Fig. 5, C and F, with Fig. 4F), the post-GO histograms in MIc became much less directional (compare Fig. 5, A and D, with Fig. 4, A and D). This further illustrates how the directionality of single-cell IDP activity is a good predictor of RTE directionality in PMd, but is a poorer predictor in MIc.

Comparison of intensity of post-GO responses in RT and DD trials

The instructional information provided by the CUE had a range of effects on the post-GO activity of PMd cells in DD trials compared with RT trials (Figs. 2, 3, and 8). The cell in Fig. 8A emitted a brisk phasic burst during the RTE of RT trials (left). In DD trials (right), a phasic early-IDP response occurred after CUE presentation followed by a sustained tonic discharge throughout the IDP. After the GO signal in DD trials, the tonic IDP discharge carried over into the RTE and ended abruptly at movement onset, but no phasic RTE response was recorded, as if the burst evoked by the GO signal in RT trials was instead evoked by the CUE signal in DD trials and not repeated after the subsequent GO signal. Therefore the cell's phasic response did not appear to be coupled temporally to the GO signal, per se, or to the motor response, but rather to the appearance of the first signal with instructional value in both trial classes. Figure 8B shows another PMd cell that also emitted a phasic post-GO response during the RTE in RT trials (left). In DD trials (right), phasic responses was emitted after both the CUE and the GO signals, neither of which was as intense as that seen after the GO signal in RT trials. The PMd cell in Fig. 8C emitted almost identical peak post-GO discharge rates in RT and DD trials, despite the presence of both early phasic and later tonic IDP activity that carried over into the early RTE of DD trials. Finally, the cell in Fig. 8D was essentially inactive in RT trials, but produced a strong ramp increase in discharge throughout the IDP of DD trials that continued into the early part of the RTE. As a result, there was an apparent increase in that cell's post-GO activity from RT to DD trials.

View larger version (49K): [in this window] [in a new window]   Fig. 8. Rasters and histograms illustrating the responses of 4 different PMd cells (A-D) in RT trials (left) and in DD trials (right). In RT trials, only the activity immediately preceding and after the GO signal is shown. In DD trials, single-trial and histogram data are shown separately oriented to both the CUE and GO signals (vertical dashed lines).

The change in post-GO activity recorded in DD trials was first assessed by comparing mean population responses at the preferred direction of all cells in each area (Fig. 9, A-C). In PMd (Fig. 9C), the population of cells showed a clear reduction in the peak magnitude of the mean post-GO response in DD (hatched histogram) as compared with RT trials (solid histogram). A similar effect was seen in MIr and MIc but to a lesser degree (Fig. 9, A and B). Figure 9, A-C, also suggested that prior information only affected the mean post-GO response during the RTE, i.e., before movement onset, in all three areas. Two effects are evident. First, the tonic activity carried over from the late IDP into the earliest part of the RTE. Second, there was an apparent reduction in the intensity of the initial phasic activity in the RTE. In contrast, the mean level and pattern of activity during the MTE and THT epochs appeared to be unchanged between RT and DD trials. For movements opposite to the preferred direction, there appeared to be little post-GO modulation (Fig. 9, D-F).

View larger version (45K): [in this window] [in a new window]   Fig. 9. A-C: population histograms of the averaged activity of contributing cells recorded at the preferred direction of each cell during the RTE in RT trials () and in DD trials () for all cells recorded in MIc (A; 152 cells), MIr (B; 72 cells), and PMd (C; 279 cells). Horizontal calibration bar, 250 ms; vertical calibration bar, 10 spikes/s. The two T-bars below each histogram to the right of the GO signal alignment indicate the mean time (±SD) of onset (left T) and end (right T) of movement. D-F: population histograms of the averaged activity of contributing cells recorded at the direction opposite to their preferred direction. G-I: scatter plots illustrating the correlation between the peak post-GO discharge intensity of each contributing cell during the time period 100-300 ms after the GO signal in RT trials (X-axis) and in DD trials (Y-axis). Dashed diagonal line, identity function; solid diagonal lines, regression functions for abscissa on ordinate, and ordinate on abscissa.

To quantify the reduction in intensity of the mean post-GO response in the RTE of DD trials, we first compared the area under the population histograms for the time interval from 100 to 300 ms after GO signal onset. The beginning of this time interval corresponded to the onset of earliest movement-related discharge in each area, while minimizing the confounding effect of the continuation of increased tonic activity from the late IDP into first part of the RTE (Fig. 9, A-C). The end of this interval corresponded roughly to the mean behavioral reaction time in RT and DD trials (Fig. 9, A-C; see METHODS). In PMd, the movement-related activity during that interval in DD trials was only 86.6% of that in RT trials. The corresponding values for MIr and MIc were 91.5 and 92.0%, respectively.

Next, to confirm these observations statistically on a single-cell basis, a sliding 100-ms window was used to determine the peak response of each cell for the time period 100-300 ms after the GO signal in RT and DD trials. In all three areas, there was a statistically significant reduction in the peak response of the cells in DD trials compared with that recorded in RT trials (Fig. 9, G-I). The mean reduction was more than twice as large in PMd (from a mean value of 48.18 to 40.24 spikes/s, 16.5% reduction; P < 0.01, paired t-test) than in either MIr (7.0% reduction; P < 0.05) or MIc (6.3% reduction; P < 0.01). Single cells in PMd also showed a much greater range of modulation of the post-GO response in DD trials compared with MIr and MIc, resulting in a lower correlation between the peak response from 100 to 300 ms post-GO between RT and DD trials in PMd than in MIr or in MIc (Fig. 9, G-I).

The degree of modification of early post-GO activity was not just a function of cortical area. The 29 movement-only cells in PMd showed peak responses that were highly correlated in RT and DD trials (r = 0.933, data not shown) with a nonsignificant difference in peak responses (P > 0.05, paired t-test). This indicated much more consistent early post-GO responses between trial classes for those particular PMd cells than for the majority of PMd cells, which were directionally tuned during the IDP of DD trials.

The sliding-window analysis was then repeated for cell activity from 300 to 1,000 ms after the GO signal, roughly corresponding to the MTE epoch and the early part of the THT (data not shown). The peak discharge in RT and DD trials was highly correlated in all three areas (MIc, r = 0.93; MIr, r = 0.96; PMd, r = 0.90), and there was no significant difference in peak discharge between RT and DD trials in all three areas (P > 0.05, paired t-test).

The preferential effect of prior information on RTE responses is also shown by the three-way ANOVA (Table 2). Most cells showed a significant main effect of direction across tasks for all three post-GO epochs. In contrast, the number of cells that showed a significant difference in responses between tasks, either as a main effect of task or a direction-task interaction, was highest during RTE then decreased abruptly in the MTE and still further during THT.

The sliding-window analysis was next used to sort the cells into three groups using, as an arbitrary criterion, a change in peak response between 100 and 300 ms post-GO of at least 20% in DD as compared with RT trials (Fig. 10). In all three areas, the largest group of cells showed similar post-GO responses in DD versus RT trials (<20% change, DD = RT, Fig. 10, D-F). This group comprised the majority of cells in MIc, decreasing in MIr and in PMd. The next largest group were cells that showed a decrease of at least 20% in the peak post-GO response in DD trials (DD < RT, Fig. 10, A-C). These cells were about twice as common in PMd as in MIc and MIr. The magnitude of the decrease in peak post-GO responses also showed a gradient across areas, going from a mean decrease of 16.1 spikes/s in MIc, to 22.0 spikes/s in MIr and 24.4 spikes/s in PMd. This resulted in a decrease in the area under the sample histogram from 100 to 300 ms post-GO by 53.3% in PMd, 27.8% in MIr, and 27.2% in MIc (Fig. 10, A-C). Finally, the smallest groups of cells showed an increase by more than 20% in their peak post-GO responses in DD trials compared with RT trials (DD > RT, Fig. 10, G-I). The numbers of cells in the three groups were statistically different (² test) among the three areas (MIc vs. MIr, P < 0.01; MIc vs. PMd, P < 0.01; MIr vs. PMd, P > 0.05).

View larger version (49K): [in this window] [in a new window]   Fig. 10. Population histograms illustrating the averaged activity of contributing cells recorded at the preferred direction of each cell during the RTE in RT trials () and in DD trials () for cells recorded in MIc (A, D, and G), MIr (B, E, and H) and PMd (C, F, and I). A-C: population histograms for those cells with a large decrease (>20%) in the peak level of post-GO discharge recorded in DD trials as compared with RT trials. D-F: population histograms for cells showing minor changes (<20%) in the peak level of post-GO discharge recorded in RT trials and DD trials. G-I: population histograms for cells with a large increase (>20%) in the peak level of post-GO discharge recorded in DD trials as compared with RT trials. Horizontal calibration bar, 500 ms. Vertical calibration bar, 10 spikes/s. n, number of cells (and percentage of the total sample in the corresponding cortical area) contributing to each histogram.

The histograms of the three groups of PMd cells also suggested that there was a relationship between the temporal pattern of activity during the IDP of DD trials, and the nature of the modulation of post-GO discharge. The PMd cells that showed a large decrease in peak post-GO responses in DD trials were most commonly those cells that emitted a brisk phasic burst confined mainly to the RTE of RT trials (Fig. 10C, ) and also a brisk short-latency early-IDP burst after the appearance of the CUE signal (Fig. 10C, ; Fig. 8, A and B), followed by modest (Fig. 8, A and B) or no (Fig. 2D) sustained tonic activity for the rest of the IDP. In contrast, cells with an increased post-GO response in DD trials often showed a relatively modest post-GO response in RT trials (Fig. 10I, ) and a pronounced tonic or incrementing ramp discharge during the IDP of DD trials (Fig. 10I, ). Further, these cells emitted relatively little or no short-latency phasic discharge after the CUE presentation (Fig. 10I; Fig. 8D). This group included most of the cells that were relatively inactive and nondirectional in the RT task but showed strong directional activity in the IDP of DD trials (Figs. 3, C and D, and 8D). As a result, much of their enhanced post-GO activity in DD trials could have been due to the continuation of the tonic IDP response into the RTE, even though we tried to compensate for this by beginning the sliding-window analysis 100 ms after the GO signal. Finally, cells that showed only minor differences in peak post-GO activity between RT and DD trials were intermediate in their discharge properties, showing either phasic or tonic IDP responses, or both (Figs. 10F, 2B, and 8C).

It was also noteworthy that the cells whose post-GO activity was similar in DD and RT trials (Fig. 10, D-F) were those with the strongest responses recorded during MTE and THT in both trial classes. This suggests that they were the cells in each area that were the most strongly coupled to the execution of the reaching movements and to active holding of the arm over the peripheral targets (Crammond and Kalaska 1996).

Modulation of IDP responses after behavioral errors

The effect of prior information on task-related activity was also shown by the effect of behavioral errors on the IDP activity in subsequent trials. Occasionally, the monkeys would make errors after the appearance of a CUE signal in DD trials, typically by failing to hold the handle within the central or peripheral windows for the required time. After an error, a trial of the same class and direction was repeated until it was successfully performed, before resuming the randomized-block sequence. Therefore after an error, the monkeys had advance information about the impending trial before it began. We frequently observed modifications of cell activity after error trials that appeared to reflect the influence of that prior information (Fig. 11).

View larger version (23K): [in this window] [in a new window]   Fig. 11. A: rasters and histograms showing the summed responses of 3 PMd cells to the appearance of the CUE signal (vertical dashed line) at the preferred direction of each cell during the IDP epoch. The left raster and histogram show data recorded in trials immediately following successful trials in any direction. The right raster and histogram show data recorded in trials immediately following unsuccessful trials in which a CUE signal appeared at the cells' preferred direction, but the monkey made an error and was not rewarded. Composite histograms were generated from 3 cells because error rates were generally so low that very few error trials occurred in a particular direction for a given cell. Horizontal calibration bar, 500 ms. Vertical calibration bar, 20 spikes/s. B: rasters and histograms showing the summed responses of 3 different PMd cells to the appearance of the CUE signal (vertical dashed line) at the preferred direction of each cell during the IDP epoch, immediately following successful trials in any direction (left) or after unsuccessful trials at the preferred direction (right).

In Fig. 11A, the raster and histogram to the left shows the combined responses of three PMd cells that emitted phasic bursts in response to the appearance of CUE signals in their preferred direction in trials immediately after successful trials in any direction. The raster and histogram to the right in Fig. 11A shows the responses of the same cells in trials in which a CUE had appeared at the preferred direction immediately following trials of the same direction in which the monkey had committed an error. The phasic CUE responses were dramatically reduced. The response reduction was dependent on a behavioral error being committed in the previous trial and was not due simply to repetition of a CUE at the same spatial location in successive trials. When we deliberately presented the same target location repeatedly regardless of behavioral result, the CUE reliably evoked a response if it was preceded by a successful trial.

Figure 11B (left) illustrates three different PMd cells that responded tonically to a CUE at the preferred direction after a successful trial. When the monkey committed an error and the same CUE was presented in the next trial (Fig. 11B, right), the elevated tonic activity initiated by the CUE in the previous trial was sustained with some abatement during the intertrial interval and began to increase further during the pre-CUE center-hold epoch. The sustained discharge presumably reflected the fact that the monkey could anticipate that the same movement direction would be required in the subsequent trial. A corresponding effect is evident for the cells with phasic responses in Fig. 11A. After the phasic burst evoked by the CUE, the tonic discharge tended to decline below the pre-CUE rate (Fig. 11A, left histogram). After errors, the reduced tonic rate tended to be sustained during the intertrial period into the subsequent trial (compare the pre-CUE tonic rate of the histogram in the right of Fig. 11A with that in the left).

Effect of nonspatial GO signals and removal of CUE signal on IDP activity

In both RT and DD trials, extinction of the central red LED and illumination of a single peripheral red LED served as the GO signal. One possible consequence of this task design is that in DD trials, the monkeys could have ignored the CUE and moved to the peripheral target identified by the single red LED. To evaluate this possibility, two monkeys were also trained to perform DD-NS and MEM trials (see METHODS). We recorded from 35 PMd and 31 MIc neurons during the performance of separate blocks of DD and RT trials, and of DD-NS and MEM trials.

Similar IDP responses were recorded in DD and DD-NS trials in both PMd (Fig. 12) and MIc. For each IDP epoch, the vast majority of cells showed no significant difference (P > 0.05) in their discharge in either direction (Table 5), even though in DD trials, the GO signal always specified the target location, whereas in DD-NS trials, it never did.

View larger version (17K): [in this window] [in a new window]   Fig. 12. Population histograms of the averaged responses of 35 PMd cells with sustained tonic discharge at the preferred direction for each cell during the IDP epoch. The activity of all 35 cells was recorded in 3 different classes of instructed-delay trials. DD, direct delay trials with spatial GO signals; DD-NS, direct delay trails with nonspatial GO signals; MEM, memorized delay trials with nonspatial GO signals. Horizontal calibration, 500 ms. Vertical calibration, 10 spikes/s. T-bars below each histogram, mean time (±SD) of movement onset.

The vast majority of cells also showed no significant difference in activity between DD-NS and MEM trials (Table 6). In particular, there was no marked change in responses in the late IDP, during which the CUE signal was still present in DD-NS trials but had been extinguished in MEM trials. As a result, the population histograms were essentially identical throughout the late stages of the IDP in both trial types (Fig. 12). These results demonstrate that sustained IDP activity was not dependent on the continued presence of a visual stimulus at the intended target location.

Oculomotor behavior

We imposed no constraints on oculomotor behavior and did not directly measure eye movements. Periodically, we watched the monkeys' eyes while collecting a cell data file. During those intermittent observations, all three monkeys displayed a similar strategy of continual random saccades between the central target and the eight peripheral targets throughout the trial. They did not appear to fixate any target for any sustained period of time at any predictable time in a trial. In particular, these highly practiced monkeys did not appear to fixate the target over which they were holding the pendulum or the location of the cues during the IDP for any extended period of time. Instead, they shifted their direction of gaze several times during the pre-CUE and IDP with occasional saccades to the central and cued targets. However, because of the lack of eye movement measurements, we cannot verify these observations quantitatively or evaluate what influence they might have had on cell activity in this study.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We compared precentral single-cell activity in both RT and DD trials. It is widely presumed that activity during the behavioral RTE of an RT task comprises components related to different putative motor planning and execution processes. The instructional cues in DD trials should initiate those planning events related to the processing of instructions, response selection, and the specification of response parameters, that can be dissociated temporally from