| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
1Centre de Recherche en Sciences Neurologiques, Département de Physiologie, Faculté de Médecine, Université de Montréal, C.P. 6128, Succursale Centre-Ville, Montreal, Quebec; and 2School of Kinesiology and Health Science, 336 Bethune College, York University, Toronto, Ontario, Canada
Submitted 27 July 2005; accepted in final form 11 February 2006
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
; Bhat and Sanes 1998; Gordon et al. 1994a,b; Krakauer et al. 2000; Morasso 1981; Soechting and Flanders 1989a,b). To execute the planned movement, those hypotheses presume that the reach plan in kinematic parameters is subsequently converted into a representation of the causal kinetics (e.g., static and dynamic forces, joint torques) necessary to determine the required muscle activity patterns. The nature of these representations and of any intervening transformations, as well as the neural populations that are involved in these operations, all continue to be the topic of vigorous debate (Andersen and Buneo 2002, 2003; Battaglia-Mayer et al. 2003; Burnod et al. 1999; Colby and Goldberg 1999; Feldman and Levin 1995; Georgopoulos and Ashe 2000; Kalaska et al. 2003; Feldman and Latash 2005; Moran and Schwartz 2000; Ostry and Feldman 2003; Rizzolatti and Luppino 2001; Todorov 2000). Nevertheless, it is evident from everyday experience that we have considerable voluntary control over all aspects of the metrics of reaching movements, including hand paths and final endpoint locations, joint rotations, arm segment motions and orientations, output forces, and muscle activity. To achieve that control, it is reasonable to presume that neuronal representations of these different movement attributes should exist in one form or another within the distributed arm motor control system.
The primary motor cortex (M1) and area 5 of the posterior parietal cortex are both involved in the control of arm movements. M1 is generally assumed to play a prominent role in generating the final cerebral cortical motor output command (Cheney and Fetz 1980; Evarts 1968, 1969; Hepp-Reymond et al. 1978, 1999; Lemon et al. 1986). In contrast, area 5 may contribute to the sensorimotor guidance of motor behavior, including the integration of multimodal sensory information and some of the early sensorimotor transformations implicated in the planning of movement (Andersen and Buneo 2002, 2003; Andersen et al. 1997; Batista et al. 1999; Battaglia-Mayer et al. 2001, 2002, 2003; Buneo et al. 2002; Caminiti et al. 1998, 1999; Colby and Goldberg 1999; Ferraina et al. 2001; Graziano and Gross 1998; Graziano et al. 2000; Kalaska 1996; Kalaska et al. 1990, 2003; Mascaro et al. 2003; Rizzolatti and Luppino 2001; Sakata 2003; Snyder et al. 1997).
Neurophysiological studies have shown that M1 activity is modulated by task kinetics, including both static and dynamic forces (Boline and Ashe 2005; Cheney and Fetz 1980; Evarts 1968, 1969; Georgopoulos et al. 1992; Gribble and Scott 2002; Hepp-Reymond et al. 1978, 1999; Kalaska et al. 1989; Lemon et al. 1986, 1998; Li et al. 2001; Maier et al. 1993; McKiernan et al. 1998; Taira et al. 1996; Thach 1978). Other studies have shown that M1 activity also covaries with the kinematics of motor output (Caminiti et al. 1990, 1991; Georgopoulos et al. 1982; Kalaska et al. 1989; Moran and Schwartz 1999; Reina et al. 2001; Schwartz et al. 1988; Scott and Kalaska 1995, 1997). By using an external load to dissociate kinetics from kinematics during a reaching task, Kalaska et al. (1989) showed that the activity of many M1 neurons was modulated both by the direction in which the arm was pulled by the external forces and by the direction of movement and the static posture of the arm during unloaded arm movements. Furthermore, the directionality of arm movement–dependent and load-dependent response modulations were strongly coupled, suggesting a functional link or even a common causal origin for those two response properties in those neurons (Kalaska et al. 1989). Nevertheless, many other M1 neurons were strongly directionally tuned for movement direction but were relatively insensitive to the presence and direction of external loads, indicating that the response properties of M1 neurons and their contributions to performance of the task were not homogeneous (Kalaska et al. 1989).
In contrast, area 5 neuron discharge is highly modulated by the direction of movement and by the posture of the arm in reaching tasks (Battaglia-Mayer et al. 2001, 2003; Buneo et al. 2002; Ferraina et al. 2001; Kalaska et al. 1983, 1990; Lacquaniti et al. 1995; Mascaro et al. 2003; Scott et al. 1997), but is much less influenced than M1 by external forces during static posture and arm movement (Kalaska and Hyde 1985; Kalaska et al. 1990). Also in contrast to M1, the directionality of movement-dependent and load-dependent response modulations of area 5 neurons were not strongly coupled (Kalaska et al. 1990). Ashe and Georgopoulos (1994) found that area 5 activity was most strongly correlated to the direction and velocity of movement but poorly correlated to movement acceleration, which is the kinematic parameter most closely correlated to movement forces through the Newtonian laws of motion. These various findings suggest that the arm movement representation in area 5 expressed primarily the kinematic attributes of motor outputs. In at least one preliminary report, however, both area 5 and M1 neurons were found to be modulated to similar degrees by the level of force exerted during an isometric task (Boline and Ashe 1998).
To re-examine the nature of motor output representations in the motor cortex, Sergio and Kalaska (1998) and Sergio et al. (2005) studied the activity of M1 neurons during two tasks specifically designed to dissociate the kinematics and kinetics of motor output. Monkeys displaced a cursor on a screen from a central target to one of eight peripheral targets by moving a weighted handle (movement task) or applying forces to a fixed handle (isometric task). The isometric task required the monkeys to generate monotonically increasing force ramps in eight directions in the horizontal plane. In contrast, the combined inertia of the arm and weighted handle in the movement task required the monkeys to transiently reverse the direction of output forces at the hand to apply a brief braking pulse to the handle to decelerate it as it approached the target. M1 neurons were strongly active in both tasks but showed changes in their response time-course that paralleled the differences in the temporal patterns of forces in the two tasks (Sergio and Kalaska 1998; Sergio et al. 2005). These results complement the previous study (Kalaska et al. 1989) in which external forces were applied at the hand during reaching movements.
In this study, we recorded the activity of area 5 neurons in the same monkeys from which motor cortex recordings had been made (Sergio and Kalaska 1998; Sergio et al. 2005). If parietal neurons are primarily processing information about movement kinematics, as found in earlier studies (Ashe and Georgopoulos 1994; Kalaska et al. 1990), area 5 neurons should be active and vary with movement direction and static postures in the arm movement task, but should not show activity that varies with the moment-to-moment changes in forces during movement. Furthermore, area 5 neurons should be significantly less active and less directionally tuned in the isometric task, which involves only minor changes in arm geometry and no displacement of the hand. Preliminary results of this study have been reported previously (Hamel-Pâquet et al. 2002, 2003).
Terms borrowed from mechanics (e.g., kinematics, kinetics, dynamics) are used throughout this paper. However, they are used here only in a general sense as convenient descriptors to express the degree to which neuronal activity covaries with the externally observable spatiotemporal form of motor outputs (task kinematics) or to their underlying causal forces, torques, and muscle activity (task kinetics), while the system is in equilibrium (statics) and while in transition between static states (dynamics). Their use does not imply that parietal cortex neuronal activity explicitly codes any particular Newtonian mechanical parameter.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
Two juvenile male rhesus monkeys (Macaca mulatta; monkey A: 3.4–6.1 kg; monkey B: 3.4–7.0 kg) were trained to perform an isometric-force task, and an arm-movement task against an inertial load (Fig. 1). In the isometric task, the manipulandum was a rigid handle attached to a 6-df force/torque transducer (F3/T10 system, Assurance Technologies) fixed at waist level 20 cm in front of the monkey. The monkey held onto the handle with its hand and used its whole arm to generate 1.5-N force ramps in different directions in the horizontal plane. In the movement task, an identical handle/force transducer assembly was installed in a box at the free end of a 1.6-m-long weighted pendulum. Extra weights were added empirically to the transducer assembly box to ensure that the ranges of dynamic and static forces were comparable in the two tasks (Sergio et al. 2005). The transducer and box weighed 1.3 kg and the pendulum itself weighed a further 1.3 kg. To localize the position of the free end of the pendulum in the horizontal (x-y) workspace of the task, the stylus of an ultrasonic digitizer was attached to the pendulum base and its spatial location was sampled at 55 Hz (0.1-mm resolution; GP-9, Science Accessories Corp.). The spatial location of the monkey's hand in the isometric task and its starting position in the movement task were identical.
|
; Sergio et al. 2005). Forces applied to the pendulum handle in the x-y plane during the movement task were measured by the force transducer installed in the handle but were not used to control the cursor position on the monitor, unlike in the isometric task. Vertical (z-axis) forces were measured in both tasks and controlled so that they remained within a small range about the horizontal plane of each task (Sergio and Kalaska 1998, 2003; Sergio et al. 2005). Behavioral tasks
A circle representing the central target first appeared on the monitor at the start of each trial (Fig. 1). The monkey generated a small static force of 0.3 N directed away from its body on the rigid handle (isometric task) or pushed the pendulum from its natural rest position to a location slightly further away from its body (movement task) to hold the cursor in the central target for a variable period of time (2,000 ± 500 ms). After this initial hold period, the central target disappeared and one of eight peripheral targets arrayed in a circle around the central target appeared. The eight targets were spaced at 45° intervals, starting from 0° at the right. The monkey generated the isometric force or the arm movement required in the horizontal plane to displace the cursor from the center to the peripheral target and held it there for 2 s to receive a liquid reward. One data file comprised 40 successful trials of one of the tasks, corresponding to five trials to each target, in a randomized-block design. The two tasks were performed in separate consecutive files of 40 trials. The order in which the tasks were performed varied from neuron to neuron. In a number of neurons, duplicate files of each task were collected to verify the stability of neuronal activity.
Four behavioral epochs were defined for the trials in both tasks. The first epoch, center-hold time (CHT), ended when the peripheral target appeared. The interval between the appearance of the peripheral target and the first significant change in force applied to the manipulandum in both tasks was called reaction time (RT). The dynamic-force time epoch (DFT) of the isometric task ended when the cursor stabilized at a constant target force direction and level. The equivalent movement-time epoch (MT) of the movement task ended when the pendulum stabilized at a constant spatial position in the peripheral target. The target-hold time (THT) epoch corresponded to the remaining period of static hold in the target circle in both tasks.
Data collection
The monkeys were implanted with a recording cylinder over area 5 of the posterior parietal cortex using standard aseptic surgical techniques and stereotaxic coordinates. These same animals were also used for a parallel study of M1 neuron activity in these same tasks (Sergio and Kalaska 1997, 1998, 2003; Sergio et al. 2005).
Conventional techniques were used to isolate and record the activity of single neurons in posterior parietal area 5 (Kalaska et al. 1989, 1990). A glass-insulated platinum-iridium microelectrode was used and advanced through the cortex to isolate neurons. Each neuron was tested in the task to see if it was active and directionally tuned in the tasks. The passive response of each neuron was then tested, if possible, to determine its peripheral input. If a neuron responded to active or passive movements of the contralateral proximal joints but not to more distal inputs, and was directionally tuned in at least one of the tasks, it was recorded and used for further analysis.
Near the end of data collection in each cylinder, small electrolytic lesions were made (5–20 µA, 10 s) in selected penetrations. When the experiment was finished, the monkeys were deeply anesthetized with barbiturates and perfused with buffered saline and formalin for monkey A or with saline, and 4% paraformaldehyde for monkey B. Pins where inserted into the cortex at known grid coordinates to delimit the cortical area from which neurons were recorded. The cortex was sectioned to permit localization of the marked penetrations.
Activity was recorded from 16 proximal arm muscles in both monkeys in recording sessions separate from the neural recordings. In each EMG recording session, two muscles were implanted percutaneously with pairs of Teflon-coated single-stranded stainless steel wires. The recorded signal was amplified (1,000–5,000 times), rectified, integrated (5-ms bins), and digitized on-line at 200 Hz.
Data analysis
An unbalanced repeated-measures ANOVA was used to test for a significant main effect of task or direction and for direction-task interactions on the mean neuron discharge rate during RT, MT, and THT epochs (P < 0.01; 5V program, BMDP Statistical Software, Los Angeles, CA). A 
2 test assessed if the frequency of effects was different between epochs (P < 0.05).
The direction-related dynamic range of neuron activity across the eight directions of motor output was determined to compare the overall effect of motor output direction on the activity of the neurons between the two tasks. For each neuron, the direction-related dynamic range was defined as the difference in mean discharge rate recorded for the two directions of motor output that showed the maximum and minimum discharge rates across the eight directions of output in a given epoch of a given task.
To compare these results between area 5 and M1, a contrast ratio analysis was performed. The contrast ratio values were calculated from the dynamic range values (DR) obtained for the movement (mvt) and the isometric (iso) tasks during each trial epoch, using the following equation
 |
The values ranged from –1 to +1 and were binned in 10 groups of 0.2 units. The two resulting distributions were then compared with each other with a Kolmogorov-Smirnov (KS) test (P < 0.05). Each distribution was also tested for a significant deviation from equal activity in each task (i.e., mean contrast ratio of 0.0; t-test, P < 0.05).
To compare the incidence of significant directional tuning of neurons between tasks, the directionally tuned neurons, determined by a bootstrap test (Crammond and Kalaska 1996, 2000; Sergio and Kalaska 1998, 2003), were counted, and a 2 test was performed. To test if the distribution of preferred directional tuning of the neurons was uniform across the population, a Rao's spacing test (P < 0.05) was done on the distribution of the preferred directions (PDs) of the sample population (Batschelet 1981).
In addition, a sliding-window analysis was performed for the isometric and movement tasks to describe the time-varying profile of the apparent instantaneous directionality of each neuron on moment to moment basis. Spike data for each trial in each task were aligned to the moment of force onset. A window of 50-ms duration was defined and advanced by steps of 10 ms, beginning 400 ms before force onset and ending 1,200 ms after force onset. The momentary mean activity of the neuron was calculated in each successive window of each trial, using whole and partial spike intervals rather than simple spike counts (Coe et al. 2002; Georgopoulos et al. 1982; Sergio and Kalaska 1998, 2003; Sergio et al. 2005). The PD of the activity was calculated in each window, using standard methods (Sergio and Kalaska 1998, 2003). A bootstrapping procedure was used to evaluate if neuron activity in each window was significantly directionally tuned with a confidence level of 95% (Crammond and Kalaska 2000; Sergio and Kalaska 2003; Sergio et al. 2005).
To evaluate the moment-to-moment net directional signal at the population level, a population-vector analysis was performed. For this analysis, momentary neuronal activity was calculated in 20-ms nonoverlapping windows, after all single-trial data were aligned to force onset. The apparent PD of a neuron can change from window to window, in part because the stochastic nature of neuron activity is accentuated when examined at such short time intervals. Furthermore, the apparent directionality of M1 neurons often changed dramatically during the MT epoch of the movement task (Sergio and Kalaska 1998; Sergio et al. 2005). This raises the question of how to identify the presumed directional influence of a given neuron on motor output at a given moment in time, and thus its contribution to the evolving population signal.
To address this question, we assumed that this directional influence on peripheral motor output remains stationary, at least over the time frame of single trials and single data files. We observed that the apparent directional tuning of muscles and single M1 neurons was generally similar between isometric and movement tasks during the RT and especially the THT epochs. Because the THT epoch is substantially longer than the RT epoch, we decided that this would give a more stable estimate of the overall directional tuning of M1 neurons. Therefore we calculated the preferred direction of the M1 neurons using the average discharge of each neuron during the THT epoch in each task and used it as the canonical PD of the neuron for the reconstruction of population vectors in the corresponding task (Sergio et al. 2005). We retained that same general approach for this study, but had to adapt it to the response patterns of parietal cortex neurons. Because parietal cortex neurons were typically most active during the dynamic (MT or DFT) phase of each task, the canonical PD of each neuron was calculated during the MT epoch of the movement task and the DFT epoch of the isometric task, rather than during THT as had been done for the M1 data (Sergio et al. 2005). Irrespective of the apparent directional tuning of the neuron in a given 20-ms time window, its presumed contribution to the population-vector signal was based on its PD during the MT or DFT of the corresponding task. Population vectors were reconstructed from the summed directional contributions of all neurons in the sample, using standard methods (Sergio et al. 2005).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
Forces were measured at the hand while the monkeys performed the movement and the isometric tasks. In the isometric task, forces were monotonically increasing ramps directed at the target that rose progressively and smoothly up to a stable force maintained throughout the remainder of the trial (Fig. 2A). In the movement task, the monkeys made corresponding monotonic ramp displacements of the arm toward targets. However, the forces applied by the hand showed a more complex "triphasic" pattern (Fig. 2B). An initial accelerative force was directed toward the targets, followed by a transient rapid reduction and change in direction of exerted force of almost 180° before the peak of velocity to slow the handle as it approached the target. This was followed by a second reversal of force direction so that it again pointed in the direction of the target. This latter force was sustained for the remainder of the trial to hold the mass of the pendulum over the target against the force of gravity.
|
Muscle activity was recorded from a total of 57 sets of EMG records collected from 16 proximal arm muscles in each monkey (see Sergio and Kalaska 2003; Sergio et al. 2005 for more details). The distribution of the preferred directions of the muscles was uniform in all epochs in both tasks (Rao's spacing test, P < 0.05). The profile of activity of muscles appeared to parallel the differences in force profiles in both tasks. For almost all the muscles studied, such as the middle deltoid (Fig. 2), the EMG activity changed monotonically during force-ramp generation and was directionally tuned in the isometric task (Fig. 2A). In the movement task, in contrast, the muscle activity showed a reciprocal "triphasic-burst" pattern in the movement task, including a brief reduction or pause in activity during movement in the directions in which the muscle was an agonist, and a delayed burst during movements in directions in which the muscle acted as an antagonist (Fig. 2B). These transient changes in EMG activity were not seen in the isometric task (Sergio and Kalaska 1998; Sergio et al. 2005). In these highly practiced animals, EMG activity was characterized by precisely timed reciprocal activation of antagonist muscles in both tasks, with little evidence of extended periods of co-activation of antagonist muscles.
Parietal area 5 neuronal activity
Neuronal activity was recorded in posterior parietal area 5 in the intraparietal sulcus of two monkeys (Fig. 3). Seventy-eight area 5 neurons that were directionally tuned in at least one task and that were related to the proximal arm were included in the sample. A large number of other task-related neurons were tested in the two tasks, but could not be held long enough to collect complete data files in both tasks, and so are not presented here. However, their properties were similar to those of the neurons reported in this paper. A few of the most rostral and medial penetrations in the left hemisphere of monkey B may have encroached on area 2 of SI. However, the cytoarchitectonic border between area 5 and area 2 is difficult to determine in cresyl-violet stained sections. The majority of neurons were recorded in penetrations made in the medial bank of the IPS, clearly in area 5 (Fig. 3).
|
), were significantly more active in the movement task than in the isometric task (Fig. 4).
|
). One neuron (1/78; 1%) showed a modest "triphasic-burst" pattern reminiscent of EMG activity (Sergio et al. 2005).
|
An ANOVA assessed the effect of task and direction on neuron discharge (Table 2). The results showed significant differences in overall discharge rate between the two tasks (main effect of task) for many area 5 neurons during RT (40/78; 51%), MT (56/78; 72%), and THT (49/78; 63%). Single-neuron discharge was also influenced by the direction of movement across tasks (main effect of direction) during RT (48/78; 62%), MT (67/78; 86%), and THT (68/78; 87%). Most strikingly, far more neurons showed differences in directional properties between the two tasks (significant task–direction interaction) in MT (61/78; 78%) and THT (62/78; 80%) than in RT (28/78; 36%). These last results in particular indicated that area 5 neurons tended to show progressively greater differences in their relation to output direction between the two tasks with time after the appearance of the output targets in the trials. For all three effects studied, the proportion of neurons showing a significant effect was different between epochs (2, P < 0.05). The analyses in the following sections provide further insight into the origin and functional implications of these ANOVA results.
|
A task-direction interaction can be caused by a difference in the gain of direction-dependent discharge modulation between the two tasks. To assess this possibility, we compared the direction-related dynamic range (see METHODS) of neuronal activity for different directions of motor output in the movement and the isometric tasks (Fig. 5) for all three trial epochs. The scatter plots show that most of the data points fell below the diagonal identity line, indicating a larger direction-related dynamic range in the movement task than in the isometric task for most neurons. The slope of the regression function was far below 1.0 (the identity line) in each trial epoch and decreased from RT (0.339) to MT (0.195) and THT (0.179).
|
|
). Whereas the distributions of ratios of area 5 neurons shifted toward +1 for all three trial epochs, the distributions for M1 neurons were slightly biased toward –1 for the RT and MT/DFT epochs (significantly different from 0.0, t-test, P < 0.05). This indicated that the dynamic ranges of M1 neurons were slightly larger on average during the isometric than the movement task for those two epochs. For the THT epoch, the M1 distribution was slightly biased toward +1, like the parietal neurons, but to a lesser degree. For all three epochs, the distributions of contrast ratios of area 5 and M1 neurons were significantly different (Kolmogorov-Smirnov test, P < 0.05), with a systematic bias toward larger positive contrast ratios (i.e., relatively greater activity in the movement task than in the isometric task) for area 5 compared with M1. Temporal analysis of single-neuron directionality
Another potential cause of a task-direction interaction is a task-dependent change in the directionality of motor output tuning of the neurons, independent of their discharge levels in the two tasks. To examine this possibility, we first compared the incidence of the directional tuning of neurons between the two tasks in each of the three epochs. More area 5 neurons were significantly directionally tuned (bootstrap test) in the movement task (RT: 33/78, 42%, MT: 66/78, 85%, THT: 62/78, 80%) than in the isometric task (RT: 11/78, 14%, MT: 34/78, 44%, THT: 35/78, 45%). The frequency of neurons that were directionally tuned was significantly different between the two tasks for each of the three epochs (2 test, P < 0.05). The distribution of the PDs of significantly directionally tuned neurons, calculated from the mean activity during the entire RT, MT, and THT epochs, was uniform in all three epochs in the isometric task and for the MT and THT epochs in the movement task (Fig. 7; Rao's spacing test, P < 0.05). During the RT epoch in the movement task, however, the distribution was nonuniform, showing a significant bimodal distribution (Rao's spacing test, P < 0.05).
|
|
|
Two population analyses were performed to see how the observations reported at the single neuron level are reflected at the population level.
In the first, population histograms were generated of the activity of all area 5 neurons that were directionally tuned in a given epoch (Fig. 10). Only the significantly directional neurons in a given epoch and task were included in the corresponding histogram. The histograms show that, in the movement task, the activity of the population of area 5 neurons increased rapidly before motor output onset and peaked at about 100 ms after output onset. The activity decreased to reach a plateau where it stayed until the end of the trial. In the isometric task, a similar pattern was observed, but the discharge rate was much lower. Note also the progressive change in the shape of the population histogram profiles when the neuronal data are aligned to the PD calculated in different epochs. The population histograms seem to be predominantly phasic in the movement task when aligned to the PD during the RT epoch, and there is relatively little difference in late tonic activity at the end of the trials in the PD and opposite direction (Fig. 10). The initial phasic component becomes less prominent and less strongly directionally tuned, and a directionally tuned late tonic discharge becomes more evident as the activity is aligned to the PD of each neuron during MT and then THT (Fig. 10). Similar trends are seen for the data in the isometric task, but the task-related responses were much smaller on average than in the movement task.
|
Note that the population histograms presented in Fig. 10 tend to exaggerate the amount of activity evoked in the area 5 population during the isometric task relative to that during the movement task, because they were generated using only those neurons that were significantly directionally tuned in a given trial epoch in the corresponding task. Far fewer neurons overall were significantly directional in the isometric task in each trial epoch. If histograms had been generated from the total sample population rather than just the directionally tuned neurons, the differences in activity level between the two tasks would have been even greater than is indicated in Fig. 10 (data not shown).
Next, a population–vector analysis was performed using the activity of the entire sample, and was compared with the average temporal profile of forces applied to the rigid manipulandum in the isometric task or the moving handle in the movement task (Fig. 11).
|
These trends were shown more clearly when the individual 20-ms population vectors were joined tip-to-tail to form neural trajectories (Georgopoulos et al. 1988). In the movement task (Fig. 11E), the neural trajectories started near the direction of desired motor output and continued along that direction for the duration of the trial, without any major inflections or reversals of direction. The same was also observed for the neural trajectories of the area 5 population in the isometric task, but the overall scale of the motor output representation was substantially smaller (Fig. 11F).
Comparison with neuronal activity in primary motor cortex in the same task conditions
Neuronal recordings were made from the caudal part of the primary motor cortex in the same tasks in these same monkeys. Those results are described in detail elsewhere (Sergio et al. 2005). The findings showed that the M1 population showed prominent temporal correlates of the differences in the time-course of output kinetics in the two tasks, including the reversal of output forces and the triphasic response patterns of many prime-mover muscles in the movement task against a heavy inertial load, that were not evident in the area 5 sample population.
To facilitate comparison of results, we present some of the M1 results here. Figure 6 compared the distributions of contrast ratios of dynamic ranges of activity from the two tasks in the two cortical areas. As already noted, there was a systematic shift in the distributions from M1, reflecting the greater activation of M1 cells in the isometric task compared with area 5. Figures 12 and 13 provide further direct contrasts of the major differences in the temporal patterns of neuronal activity in the two cortical areas during the movement task.
|
|
). The area 5 population (Fig. 12, thick lines) emitted a single phasic response before and during movements in the preferred direction of each neuron and relatively little change in mean activity for movements in the opposite direction. In contrast, the corresponding M1 population showed an early phasic burst peaking at movement onset in the preferred direction (Fig. 12, thin lines), followed by a brief decrease in mean activity and a second increase in activity. In the opposite direction, the M1 population showed an initial decrease in activity followed by a brief strong phasic burst preceding and during the deceleration phase of the movements. This delayed burst was completely absent in the area 5 population (Fig. 12, thick lines).
Figure 13 shows the temporal evolution of the distribution of directionally tuned activity of single cells and the net resultant population vector at 40-ms time intervals during the movement task in area 5 (cf. Fig. 11) and in M1, during arm movements in one direction, 180° (modified from Fig. 10B of Sergio et al. 2005). The area 5 population became directionally tuned before movement onset. Although the overall magnitude of the net population signal varied with time, its direction remained relatively constant and oriented near 180° at all times after the population became directionally active. In contrast, the net directional signal in M1 initially pointed near the direction of desired movement, but diminished and reversed in direction during the time period from 
200 to 400 ms after movement onset, before reversing direction again to point back in the desired direction of movement.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
One notable finding of this study is that the presence of a large inertial load did not significantly alter the overall relation of area 5 activity to arm movement kinematics. The present sample population showed the same broad tuning as a function of the direction of movement and the final arm postures that has been described in many earlier studies (Battaglia-Mayer et al. 2001, 2003; Buneo et al. 2002; Caminiti et al. 1998; Ferraina et al. 2001; Georgopoulos et al. 1984; Kalaska and Crammond 1995; Kalaska and Hyde 1985; Kalaska et al. 1983, 1990; Lacquaniti et al. 1995; Scott et al. 1997). This shows that the underlying movement-related tuning functions of the cells were robust to the perturbation imposed by the inertial loads.
There were two principal findings in this study. First, area 5 neurons were typically more active, and often exclusively active, in the movement task than in the isometric task, even though the latter required similar overall levels of hand-centered output forces and EMG activity as in the movement task. Nearly one-half of the sample population (45%) were active and directionally tuned in the movement task but relatively or completely inactive in the isometric task. Second, the temporal pattern of area 5 activity in the movement task did not reflect the time course of the forces and muscle activity during the arm movements. In the movement task, the temporal pattern of output forces at the hand showed an initial rapid increase of force exerted against the task handle in the desired direction of movement, followed by a transient reversal of forces to decelerate the motion of the handle as the monkey approached the target. The transient reversal of forces was also clearly reflected in the muscle activity, which typically showed the classic reciprocal triphasic-burst pattern during arm movement. Area 5 activity, at both the single-neuron and population level, showed little or no evidence of the transient reversal of the directionality of output forces and muscle activity during movement. The preferred direction of each neuron, if it was directionally tuned in the movement task, tended to remain relatively constant throughout the trial, as shown by a sliding-window analysis. Only 1 of the 78 area 5 neurons showed a weak "triphasic" response pattern in the movement task. Furthermore, a population–vector analysis of the activity of the sample population showed that the net directional signal generated by the population remained oriented in the direction of movement throughout its duration. In summary, area 5 neurons were typically much more active in the movement task than the isometric task, and there was no correlate at either the single-neuron or population level of the transient braking pulse of forces that is a prominent feature of the time course of motor output in the movement task.
These results complement an earlier study of parietal area 5 (Kalaska and Hyde 1985; Kalaska et al. 1990). Kalaska and Hyde (1985) showed that constant external loads applied to the arm in different directions had only a modest effect on area 5 activity when monkeys actively held their arm in a fixed posture against the loads, compared with the much greater sensitivity of M1 neurons to the same loads. Similarly, the movement-related activity of the area 5 neurons was relatively unaffected at the single-neuron and particularly at the population level, when the monkeys made reaching movements along constant spatial hand paths while the arm was pulled in different directions by the external loads (Kalaska et al. 1990). The earlier and present studies show that parietal area 5 activity is far less modulated by constant loads during active postural maintenance, by constant or inertial loads during arm movements, and by isometric force generation than it is by arm movements and by different arm postures themselves. In other words, area 5 activity showed far weaker correlations with the time course and directionality of static and dynamic output forces and muscle activity (task kinetics) across those different task conditions than it did with the spatiotemporal form (task kinematics) of the arm motor output.
Ideally, if the arm movement representation in area 5 only signaled properties of the spatiotemporal form of the output and not about its causal forces and muscle activity, one would expect all area 5 neurons to be inactive during the isometric task. However, about one-half of the neurons were modestly active and directionally tuned in the isometric task when no limb movements are presumably being planned or executed. A number of possible explanations related to peripheral motor events or to central processes could account for this activity.
For instance, it could partly reflect the fact that the hand and the arm are not completely rigid, and that small movements and postural changes of the hand, arm, shoulder or trunk occurred while the monkeys generated the isometric forces in different directions. An indeterminate amount of the activity during the isometric task could reflect proprioceptive signals about the kinematics of these uncontrolled small motions and posture changes. Furthermore, ascending proprioceptive inputs to the primary somatosensory cortex (S1) do not signal purely kinematic information only about arm movements and changes in posture. On the contrary, the activity of many S1 neurons is also modulated by the changes in the forces and muscle activity required to compensate for external loads during arm movements or to generate isometric output forces (Fromm and Evarts 1982; Jennings et al. 1983; Prud'Homme et al. 1994; Wannier et al. 1991). Part of the area 5 activity in the isometric task could have been a residual trace of any kinetics-related information encoded in the peripheral afferent input, and relayed into area 5 via cortico-cortical projections from S1. This may have been compounded by the possibility that some of the recorded neurons may have been in the caudal part of S1 (area 2) itself, rather than in area 5 proper (Fig. 3).
Alternatively, the activity in the isometric task might be of central origin. For instance, it could be an efference copy of the isometric motor output command that is used in area 5 to deconvolve kinematics and kinetics signals, or to contribute to on-line error-correction mechanisms or other possible functions (Kalaska 1996; Kalaska et al. 2003). It could represent an abstract central signal related to the behavioral goal to generate a directional output to displace the cursor to the targets, independent of the causal kinetics required to accomplish the goal, and even though no physical displacement of the arm is intended in the isometric task (Crammond and Kalaska 1989; Ferraina and Bianchi 1994; Kalaska and Crammond 1995; Snyder et al. 1997). Finally, part of the isometric task-related modulation of area 5 neurons may have been caused by visual inputs about cursor motions or target locations or signals about the direction of gaze as the monkeys observed task events displayed on the computer monitor during the dynamic phases of each task, rather than to arm-related output (Batista et al. 1999; Battaglia-Mayer et al. 2001; Buneo et al. 2002; Ferraina et al. 2001; Graziano et al. 2000). However, visual and oculomotor influences have been documented primarily in parts of the superior parietal cortex that are located medial to the recording sites in this study.
Any combination of these peripheral and central factors could have contributed to the area 5 activity during the isometric task. We suggest that this activity is largely nonfunctional biological noise in a predominantly kinematic representation of motor behavior, but we cannot categorically reject the possibility that it indicates a role for area 5 in the control of motor output in an isometric-force task or has some other undetermined function.
Comparison of neuronal response properties between parietal area 5 and primary motor cortex
The behavior of the parietal area 5 neurons in this study is in sharp contrast to the activity of neurons recorded in the caudal half of the primary motor cortex (M1) of the same monkeys. M1 neurons were typically as active and often more active in the isometric task than in the movement task (Sergio and Kalaska 1997, 1998
