Visuomotor Processing as Reflected in the Directional Discharge of
Premotor and Primary Motor Cortex Neurons
M.T.V. JOHNSON, 1 J. D. COLTZ, 3 M. C. HAGEN, 3 AND T. J. EBNER 1 – 3
1
Departments of Neuroscience and Neurosurgery and 2 Department of Physiology, and the 3Graduate Program in
Neuroscience, University of Minnesota, Minneapolis, Minnesota 55455
The costs of publication of this article were defrayed in part by the
payment of page charges. The article must therefore be hereby marked
‘‘advertisement’’ in accordance with 18 U.S.C. Section 1734 solely to
indicate this fact.
direction, providing a neural substrate for nonstandard stimulusresponse mappings.
INTRODUCTION
The dorsal premotor cortex (PMd) has a hypothesized
role in the transformation of sensory stimuli into appropriate
motor behaviors (Crammond and Kalaska 1996; Godschalk
et al. 1981; Johnson et al. 1996b; Mushiake et al. 1991;
Weinrich and Wise 1982; Wise et al. 1992, 1997). Particular
emphasis has been placed on visuomotor transformations,
hypothesized to involve pathways from extrastriate visual
areas to PMd and the primary motor cortex (MI) via the
superior parietal lobule (for reviews see Johnson et al.
1996b; Milner and Goodale 1995; Wise et al. 1997). Neuronal populations in and surrounding the superior parietal
lobule are directionally modulated before and during reaching movements (Johnson et al. 1996b; Kalaska et al. 1983),
by visual and somatosensory stimulation (Albright 1984;
Colby and Duhamel 1991; Steinmetz et al. 1987), and during
conditional sensorimotor behavior (Mountcastle et al.
1975). Thus a variety of visuomotor signals reach the PMd.
The discharge of PMd and MI neurons has a pronounced
sensorimotor character. In instructed delay tasks, PMd neurons exhibit prominent and prolonged premovement activity.
This premovement activity is directionally tuned and modulated by the visual or remembered direction of the target
(di Pellegrino and Wise 1993; Wise et al. 1992). Similar
premovement discharge has been described in MI (Alexander and Crutcher 1990; Crutcher and Alexander 1990; Georgopoulos et al. 1989a; Johnson et al. 1996b; Tanji and Evarts
1976). Premovement activity in the PMd is highly modulated by visuospatial information that is significant for upcoming motor acts (Crammond and Kalaska 1994; di Pellegrino and Wise 1993; Johnson et al. 1996b; Shen and Alexander 1997a,b; Zhang et al. 1997). In addition, gaze
(Boussaoud 1995) and spatial attention (Boussaoud and
Wise 1993; di Pellegrino and Wise 1993) modulate the discharge of PMd neurons, as does motor error (Flament et al.
1993). The discharge of PMd and MI neurons is also highly
correlated with several parameters of movement including
direction, position, amplitude, speed, and force (for reviews
see Ashe 1997; Georgopoulos 1986; Wise et al. 1997).
How does this transformation of a visual target into a
motor signal occur? Several lines of evidence point to a
distributed process by which the visuospatial activity is processed more rostrally (i.e., PMd) and the motor commands
more caudally (i.e., MI). Premovement neuronal activity
decreases along a rostral-caudal gradient from the PMd to
0022-3077/98 $5.00 Copyright q 1999 The American Physiological Society
/ 9k2f$$de41
J278-8
01-26-99 09:03:21
neupal
LP-Neurophys
875
Downloaded from http://jn.physiology.org/ by 10.220.32.246 on July 31, 2017
Johnson, M.T.V., J. D. Coltz, M. C. Hagen, and T. J. Ebner.
Visuomotor processing as reflected in the directional discharge of
premotor and primary motor cortex neurons. J. Neurophysiol. 81:
875–894, 1999. Premotor and primary motor cortical neuronal
firing was studied in two monkeys during an instructed delay,
pursuit tracking task. The task included a premovement ‘‘cue period,’’ during which the target was presented at the periphery of
the workspace and moved to the center of the workspace along
one of eight directions at one of four constant speeds. The ‘‘track
period’’ consisted of a visually guided, error-constrained arm
movement during which the animal tracked the target as it moved
from the central start box along a line to the opposite periphery
of the workspace. Behaviorally, the animals tracked the required
directions and speeds with highly constrained trajectories. The eye
movements consisted of saccades to the target at the onset of the
cue period, followed by smooth pursuit intermingled with saccades
throughout the cue and track periods. Initially, an analysis of variance (ANOVA) was used to test for direction and period effects
in the firing. Subsequently, a linear regression analysis was used
to fit the average firing from the cue and track periods to a cosine
model. Directional tuning as determined by a significant fit to the
cosine model was a prominent feature of the discharge during both
the cue and track periods. However, the directional tuning of the
firing of a single cell was not always constant across the cue and
track periods. Approximately one-half of the neurons had differences in their preferred directions (PDs) of ú457 between cue and
track periods. The PD in the cue or track period was not dependent
on the target speed. A second linear regression analysis based on
calculation of the preferred direction in 20-ms bins (i.e., the PD
trajectory) was used to examine on a finer time scale the temporal
evolution of this change in directional tuning. The PD trajectories
in the cue period were not straight but instead rotated over the
workspace to align with the track period PD. Both clockwise and
counterclockwise rotations occurred. The PD trajectories were relatively straight during most of the track period. The rotation and
eventual convergence of the PD trajectories in the cue period to
the preferred direction of the track period may reflect the transformation of visual information into motor commands. The widely
dispersed PD trajectories in the cue period would allow targets to
be detected over a wide spatial aperture. The convergence of the
PD trajectories occurring at the cue-track transition may serve as
a ‘‘Go’’ signal to move that was not explicitly supplied by the
paradigm. Furthermore, the rotation and convergence of the PD
trajectories may provide a mechanism for nonstandard mapping.
Standard mapping refers to a sensorimotor transformation in which
the stimulus is the object of the reach. Nonstandard mapping is
the mapping of an arbitrary stimulus into an arbitrary movement.
The shifts in the PD may allow relevant visual information from
any direction to be transformed into an appropriate movement
876
M.T.V. JOHNSON, J. D. COLTZ, M. C. HAGEN, AND T. J. EBNER
/ 9k2f$$de41
J278-8
speed and direction of the upcoming movement, 2) elicit
controlled movements of a specified speed and direction in
which both parameters were controlled independently, and
3) require continual matching of visual input with motor
output to maintain a straight trajectory within an error envelope. Analysis of the directional tuning of the discharge of
these neurons (Caminiti et al. 1991; Fu et al. 1993; Georgopoulos et al. 1982; Schwartz 1992) was used to study the
neuronal transformation of a dynamic sensory stimulus into
a motor output. These results reveal that the directional tuning is not constant across the cue and track periods and
that during the cue period the preferred direction gradually
changes. A preliminary account of these findings has been
presented (Johnson et al. 1996a).
METHODS
Behavioral task
Experimentation was conducted according to the Guiding Principles for Research Involving Animals and Human Beings as endorsed by the American Physiological Society and was approved
by the Institutional Animal Care and Use Committee of the University of Minnesota. Two female rhesus monkeys (Macaca mulatta,
4–6 kg) used a two-joint manipulandum to make visually guided
arm tracking movements in the horizontal plane (Fu et al. 1993;
Ojakangas and Ebner 1992). The task was performed on a workspace based on a modified center-out, eight target design (Georgopoulos et al. 1982), and its trial sequence is schematized in Fig.
1. The movements of the hand/manipulandum were displayed as
a cross-shaped cursor on a computer monitor that was placed vertically, Ç50 cm in front of the animal. The center of the monitor
was aligned with the animal’s sagittal plane, as was the center of
the task workspace. The dimensions and size of the screen workspace and targets were identical to those of the actual physical
workspace; however, horizontal movements were guided by visual
cues mapped on a vertical plane. The task required the animal to
track square targets that moved at constant speeds on the screen.
The task sequence began by the animal positioning the center of
the cursor (0.5 cm diam) in a square start box (1.44 cm2 ) at the
center of the workspace for a random period ranging 1–2 s. After
this hold period, a square target box of the same size appeared at
the periphery of the workspace at one of eight target positions (0–
3157 in 457 increments) at a distance of 5 cm. This target then
moved at one of four constant speeds (2, 3, 4, and 5 cm/s) along
a straight line toward the start box. The time between target appearance and intersection with the start box is the ‘‘cue’’ period, and
is analogous to the instructed delay period of other paradigms (for
example, Crammond and Kalaska 1994; Georgopoulos et al. 1989a;
Johnson et al. 1996b). During the cue period, the animal was
required to maintain the center of the cursor within the start box.
The animal could view its hand and manipulandum, but the
tracking requirements of the task demanded that the animal concentrate on the screen.
When the cue target reached the central start position, the start
box was extinguished, and the target continued to move along the
same trajectory at the same speed for 5 cm. During this interval,
called the ‘‘track’’ period, the animal was required to make an arm
tracking movement at one of four constant speeds (2, 3, 4, 5 cm/
s) for a distance of 5 cm. At all times during the track period, the
animal was required to maintain the center of the cursor entirely
within the confines of the target box. At any point in the trial
sequence, deviation of the cursor from the target aborted the trial
and initiated a new trial sequence. All successful trials were followed by a juice reward, and after an interval of 4–8 s, a new
trial was initiated. This trial sequence was repeated over the eight
directions and four target speeds in a randomized block design,
01-26-99 09:03:21
neupal
LP-Neurophys
Downloaded from http://jn.physiology.org/ by 10.220.32.246 on July 31, 2017
MI, corresponding to a topographically organized input from
the superior parietal lobule (Alexander and Crutcher 1990;
Johnson et al. 1996b). There is also a rostral-caudal gradient
from PMd to MI reflecting the processing of visuospatial to
motor information (Shen and Alexander 1997a,b). Evidence
also points to a temporal sequencing of the visuomotor transformation in which the more visuospatial aspects of the task
are processed earlier and the more motoric features later
(Shen and Alexander 1997a,b; Zhang et al. 1997). For example, in tasks in which the visual information is dissociated
from the actual movement, the discharge is characterized by
an early phase reflecting the direction of the visual cue and a
late phase reflecting the direction of the upcoming movement
(Crammond and Kalaska 1994; Georgopoulos et al. 1989b;
Lurito et al. 1991; Shen and Alexander 1997a,b; Wise et al.
1996; Zhang et al. 1997). The population vector has been
shown to rotate from the direction of a visual cue to that of
the upcoming movement (Georgopoulos et al. 1989b; Lurito
et al. 1991). Also, there is sequential processing of information about the direction, position, and amplitude of a movement in the discharge of PMd and MI neurons (Fu et al.
1993, 1995). Therefore, in PMd and MI, both spatial gradients and temporal sequencing of information appear to contribute to the transformation of a visual stimulus into a limb
movement.
A more specific hypothesis states that the role of the PMd
in visuomotor transformations is nonstandard mapping
(Wise et al. 1996). For reaching movements, standard mapping refers to a sensorimotor transformation in which the
stimulus is the object of the reach. Nonstandard mapping
is the mapping of an arbitrary stimulus onto an arbitrary
movement. The results of imaging, lesioning, and electrophysiological studies support the hypothesis that the PMd
plays an important role in nonstandard mapping (for review
see Wise et al. 1996). How might nonstandard mapping
be achieved? Evidence exists for the learning of complex
stimulus-response relationships in the PMd (Boussaoud and
Wise 1993; di Pellegrino and Wise 1993). However, learning of all possible stimulus-movement combinations may be
computationally prohibitive, and it is possible that neural
substrates exist for the more common types of nonstandard
mappings. Furthermore, the brain may not distinguish between standard and nonstandard mapping and may use a
common neural mechanism for both. For example, consider
the processing of movement direction during a visually
guided, instructed-delay reaching task. If standard mapping
was operative, the prediction is that directional tuning would
be constant throughout the task (Wise et al. 1996). Another
possibility is that the directional tuning changes throughout
the task, consistent with a nonstandard mapping process in
which a visual stimulus from one direction can be transformed into movements of a different direction.
To address this question the discharge of PMd and MI
neurons was analyzed while monkeys performed a two-dimensional visually guided tracking task with an instructed
delay period. Most previous studies have examined motor
cortical activity during steplike reaching movements (for
review see Ashe 1997; Georgopoulos 1986; Wise et al.
1997). In this report a new paradigm based on two-dimensional pursuit tracking with the arm was used. The paradigm
was designed to 1) provide a long instructed delay period
that continuously presented visual information about the
VISUOMOTOR PROCESSING IN MOTOR CORTICAL NEURONS
and the discharge of each neuron was evaluated over 10 trials for
each direction-speed combination. The cue and track periods were
equal in duration, with each period lasting 2.5, 1.67, 1.25, and 1
s for the target speeds of 2, 3, 4, and 5 cm/s, respectively.
Surgical, electrophysiological, and histological procedures
Once the monkeys had learned the tracking task, a chronic recording chamber (19 mm ID) was stereotaxically placed, and a
head fixation halo was implanted (for details see Fu et al. 1993).
The chamber was positioned to straddle the premotor and primary
motor cortices. Anesthesia was induced and maintained with ketamine (20 mgrkg 01rh 01 im) and xylazine (1–2 mgrkg 01rh 01 im)
during the aseptic surgical procedure. Postoperatively, buprenorphine (0.05 mg/kg im) was administered for 5 days. Chambers
/ 9k2f$$de41
J278-8
were placed over three cortical hemispheres of the two monkeys
(right and left in animal CR and left in DN).
After the animal’s recovery, extracellular single-unit recording
began using paralyene-coated tungsten microelectrodes (3–10
MV ) and conventional electrophysiological techniques (Fu et al.
1993). A neuron was recorded if its activity was audibly modulated
during the performance of the task or related to active reaching
movements. After isolation of a single unit, the action potentials
were time-amplitude discriminated. The output of the discriminator
was digitized and stored at 1-ms intervals. The digitized data were
analyzed in bin sizes of 20 ms and averaged over the number of
trials. These binned firing frequencies were used in the period
regressions of cell discharge as a function of direction. For contour
plots of cell activity and calculations of a cell’s preferred direction
trajectory (see Analysis), each bin was convolved with a Gaussian
function with a standard deviation of 20 ms.
The animal’s tracking movements were calculated by recording
the time-varying position of the manipulandum. The x and y positions of the manipulandum were calculated geometrically using
angular data sampled from the output of the manipulandum potentiometers. The position data were sampled at 1 kHz and smoothed
with the use of a 21-point moving average to prepare the position
data for numeric differentiation. Differentiation yielded the x,y
velocities from which the tangential velocity was computed. Position and velocity data were then compressed into 20-ms bins. Averaging of the position and velocity data over the 10 trials was
performed after smoothing, differentiation, and compression. Eye
movements were recorded using an infrared oculometer (Bouis
Instruments, Karlsruhe, Germany). The vertical and horizontal position data were digitized at 200 samples/s and also compressed
into 20-ms bins, but not smoothed. Electromyographic (EMG)
activity was acquired using percutaneously inserted intramuscular
wire electrodes. The signal was amplified, filtered (10–1,000 Hz
band-pass), and digitized at 2,000 samples/s. The EMG signal
was digitally rectified, compressed into 20-ms bins, and averaged
over 10 trials. Eight muscles were recorded in both monkeys: pectoralis major (n Å 7 records), spinal deltoid (n Å 7), clavicular
deltoid (n Å 9), biceps (n Å 10), triceps (long and lateral heads,
n Å 17), extensor carpi radialis (n Å 11), and flexor carpi ulnaris
(n Å 12). Both the eye movement and EMG recordings were
obtained during experimental sessions in which single-unit recordings were not undertaken.
After the completion of all electrophysiological and behavioral
recordings, various positions in the chamber were marked by electrolytic lesions. Then each animal was initially anesthetized with
ketamine (20 mg/kg im) and xylazine (1 mg/kg im), followed by
a lethal dose of pentobarbital sodium (150 mg/kg ip). Intracardiac
perfusion with saline containing heparin was followed by perfusion
with Zamboni’s fixative (4.3 g NaOH, 20 g paraformaldehyde,
18.8 g NaH2PO4 , and 150 ml picric acid in 850 ml H2O). After
removal of the brain, 50-mm frozen sections perpendicular to the
longitudinal cerebral fissure were cut on a microtome throughout
the areas of interest and stained with thionin. Recording tracks
and electrolytic lesions were identified to map the microelectrode
penetrations. The number of large ( ú29 mm) pyramidal cells in
layer V was determined in each section to estimate the boundary
between the premotor and primary motor cortex and for comparison of the recording locations with previous reports (Dum and
Strick 1991; Fu et al. 1993; Weinrich and Wise 1982). Specifically,
the number of large pyramidal cells was averaged along the cortical
surface of each slice in 3-mm increments. Density maps of the
large pyramidal cells across the cortical surface were constructed,
and the isodensity contour line of eight large pyramidal cells was
taken as an estimate of the boundary between PMd and MI.
Analysis
Neuronal spike trains and kinematic trajectories were aligned on
the target onset at the start of the cue period, before averaging.
01-26-99 09:03:21
neupal
LP-Neurophys
Downloaded from http://jn.physiology.org/ by 10.220.32.246 on July 31, 2017
FIG . 1. Schematic of the pursuit tracking task. Cue and track periods
are shown. Arrow denotes the direction of the constant speed target. Plus
sign shows the location of the cursor through the trial sequence. See METHODS for further details.
877
878
M.T.V. JOHNSON, J. D. COLTZ, M. C. HAGEN, AND T. J. EBNER
f ( u ) Å b0 / b1 cos ( u ) / b2 sin ( u )
(1)
Alternatively, this can be expressed as
f ( u ) Å b0 / c1 cos ( u 0 upd )
(2)
The peak of the cosine function, referred to as the preferred
direction (PD), is given as upd in Eq. 2 (for further details see
Georgopoulos et al. 1982, 1984; Mardia 1972). Regression coefficients b0 and c1 denote the mean discharge over all directions and
the modulation of the firing as a function of direction, respectively.
Goodness-of-fit of the data to the cosine model (Eq. 1) was expressed as R 2 , with model significance (P õ 0.05) requiring R 2 ú
0.7. The depth of directional modulation, Idir , was calculated by
normalizing the slope of the regression, c1 , to the averaged firing
over all directions, b0 (Georgopoulos et al. 1982, 1988). Neurons
whose discharge fit significantly to the cosine model will be referred to as ‘‘directionally tuned.’’
The PD was referenced to the direction of the moving target for
all analyses. For example, in the diagram of the paradigm shown
in Fig. 1, the direction of the moving target is 457 for both the cue
and track periods. Therefore, if a cell’s discharge during the cue
period was highest for targets moving from the periphery at 2257
toward 457, it was referred to as having a PD of 457. Similarly, if
a cell’s discharge during the track period was highest for movements from the center of the workspace out toward 457, the PD
was defined as 457.
For further analysis of the directional tuning, the cue and track
periods were halved into early and late epochs. Equivalent analysis
of the PD of discharge for each of these smaller epochs was performed. To increase the temporal resolution of the directional tuning, cosine regressions (Eq. 1) were also performed over sequential
20-ms bins (Mason et al. 1998). This provided a nearly continuous
estimate of the PD as a function of time and was referred to as the
‘‘PD trajectory.’’ Here, discharge in each time bin was evaluated
for a significant fit using a criterion of R 2 ú 0.7. An additional
criterion was required of the PD trajectories to ensure that the
determination of the PDs based on 20-ms bins reflected the directional tuning as a function of time. The depth of modulation (Idir )
for each bin was required to be characteristic of the Idir obtained
from the tuning curves based on the cue or track period averages.
The Idir for each bin was required to be greater than the Idir –1 SD
/ 9k2f$$de41
J278-8
obtained from the entire cue or track period. The 5–95% confidence interval for the location of the PD for each 20-ms bin was
also calculated based on the von Mises distribution (Batschelet
1981).
As shown in RESULTS the PD trajectories had a pronounced
rotational character. To characterize this property of the PD trajectories, the latency, duration, and rotational velocity were calculated
for the neurons demonstrating significant cue and track period PDs.
Latency to directional discharge was defined as the time at which
three contiguous 20-ms bins demonstrated significant directional
tuning. Latency values were restricted to 1,000 ms to allow for
comparison over the different cue period lengths. The duration of
significant directional discharge was defined as the number of bins
with significant directional tuning, expressed as a percentage of
the total trial duration. The instantaneous rotational velocity of the
PD trajectory was calculated between contiguous significant bins.
Also, the maximal and mean rotational velocities were determined
for each neuron and averaged over the population.
The EMG activity in the cue and track periods was also analyzed
for significant directional effects using the single trial ANOVA
(P õ 0.01, Tukey multiple comparisons), similar to that undertaken for the cell discharge. The timing of the EMG activity was
estimated using a binwise procedure for detecting a significant
change in the EMG level, defined as exceeding the mean hold
period activity by {3 SD.
RESULTS
Characteristics of the tracking task performance
Tangential velocity profiles and hand trajectories resulting
from the tracking of the four constant speed targets are
shown in Fig. 2. During the cue period, there were no systematic excursions in the hand velocity at any target speed.
For the slowest 2-cm/s target speed trials, there were occasional increases in hand velocity while in the start box 100–
200 ms before the onset of the track period. The hand velocity increased over the initial 200–400 ms of the track period
to approach the target velocity. The initial rise in hand velocity was followed by an overshoot and several subsequent
crossings of the target velocity. Small fluctuations in the
hand velocity were present at all target speeds, but the paradigm limited these excursions to within the target area, otherwise the trial was aborted. The slight irregularities in the
hand velocity reflect frequent feedback-dependent error corrections required for visually guided pursuit tracking. For
the fastest 5-cm/s target speed trials, hand velocity could
increase Ç100–200 ms before the complete superposition
of the start and target boxes. The hand velocity climbed to
a value of 5 cm/s over the initial 300–400 ms of the track
period and then fluctuated around the target velocity. Further
examples of the tangential hand velocity are shown superimposed on the firing data in Figs. 6–8. The tracking trajectories for the eight directions were remarkably straight and
similar for the four different target speeds. This was due to
the fact that the paradigm constrained both speed and path.
There was little directionally tuned EMG activity during
the cue period in the eight arm muscles tested. The patterns
of the arm and shoulder EMG are shown in Fig. 3 for two
shoulder muscles (pectoralis major and posterior deltoid),
an elbow extensor (long head of the triceps), and a wrist
flexor (flexor carpi ulnaris). In the cue period the directional
modulation in EMG activity was extremely small to nonexistent, a result also found for other instructed delay tasks
(Georgopoulos et al. 1989a). ANOVA demonstrated no sig-
01-26-99 09:03:21
neupal
LP-Neurophys
Downloaded from http://jn.physiology.org/ by 10.220.32.246 on July 31, 2017
Firing and kinematic data were then averaged over 10 repetitions
for each of the 8 directions over the 4 target speeds. Neuronal firing
frequency histograms with 20-ms bin increments were constructed
from the averages. The firing was analyzed over three time intervals: the hold, cue, and track periods. The ‘‘hold’’ period represented the 500 ms before the appearance of the cue target, corresponding to the time in which the monkey held the cursor in the
center start box. In addition to determining the mean firing from
the trial averages, the standard deviation of the firing from the
individual trial was calculated for the different periods.
Analysis of variance (ANOVA) based on the firing in individual
trials was used as the initial step in assessing task-related modulation (Alexander and Crutcher 1990; Crammond and Kalaska 1996;
di Pellegrino and Wise 1993; Georgopoulos et al. 1982; Shen and
Alexander 1997a). The ANOVA was used to determine whether
direction had a significant effect on the discharge in the cue or
track periods and to detect significant changes in firing over the
trial sequence from the hold, cue, to track periods. The criterion
for either a direction or period effect was P õ 0.01 (Tukey multiple
comparisons). A neuron was considered task related if direction
or period had a significant effect on its discharge.
Having identified the cells with a significant direction effect
using the ANOVA, regression analyses were used to determine
whether the discharge could be fit to a cosine model (Georgopoulos
et al. 1982; Mardia 1972). For each target speed, the mean firing
rate ( f ) for each cue or track direction ( u ), was fitted to a cosine
function
VISUOMOTOR PROCESSING IN MOTOR CORTICAL NEURONS
nificant direction effect for the cue period (P ú 0.01) for
all eight muscles sampled (n Å 73 records). As expected,
the EMG activity was highly modulated during the track
period in all muscles (P õ 0.01). The ANOVA was based
on the EMG activity in the entire cue period and may have
missed changes in EMG activity that occurred before the
onset of movement. Therefore an additional binwise procedure was performed to detect the timing of any changes
in EMG activity. A significant change was defined as that
exceeding the mean hold period activity by {3 SD. The
times of significant change invariably occurred at the cuetrack period transitions (Fig. 3). Over all directions and
muscles, the binwise procedure detected significant EMG
changes at an average of 37 { 71 and 112 { 124 ms before
track onset for the 5- and 2-cm/s trials, respectively. There
were no significant differences in the onset of the EMG
changes across the eight muscles recorded (P ú 0.1,
ANOVA). Therefore, as shown previously (Turner et al.
1995), the EMG activity during the cue period was modulated
at most for a brief period ( Ç100 ms) before track onset.
/ 9k2f$$de41
J278-8
Eye movements, although unconstrained by the experimental paradigm, were also measured. Figure 4 details the
horizontal and vertical position traces and the workspace
trajectories of the left eye for four of the eight directions of
arm tracking. For the slowest target speed of 2 cm/s, the
eye tracking behavior may be summarized by a series of
five stages, best appreciated in the time plots of horizontal
(x) and vertical (y) eye position. First, before the appearance
of the target while the hand was in the start box, saccades
subtending up to 10–207 were directed over the entire workspace, apparently as the animal searched for the target. The
saccades covered an area of the workspace larger than the
10 cm diam of the circular target array. Second, Ç100–200
ms after the appearance of the target, the saccades stopped
and the moving target was smoothly pursued for 400–500
ms. Third, after this period of smooth pursuit of the target,
a second phase of saccades occurred, lasting Ç1,000–1,200
ms. It should be stressed that in some trials the animal maintained smooth pursuit throughout the entire cue period and
that this period of saccadic eye movement was highly dependent on the target speed. Fourth, Ç700–900 ms before the
track period, smooth pursuit of the target was resumed and
continued through the first half of the track period. Fifth,
during the final portion of the tracking period, the saccades
resumed.
The faster target speed of 5 cm/s resulted in a simplified
eye movement behavior, summarized as three stages through
the trial sequence (Fig. 4). As for slower tracking, the monkey initially made large amplitude saccades during the hold
period until Ç200 ms into the cue period. This was followed
by the eyes moving to the target and smooth pursuit. The
smooth pursuit continued through the cue period well into
the track period, at which point the saccades resumed. For
the faster target velocities, few saccades occurred during the
cue period. Apparently, there was not sufficient time to allow
for saccades in the cue period and still perform the task at
the highest tracking velocities.
Directional modulation of neural activity
Neuronal activity was recorded in two animals, CR and
DN. Of the total 240 neurons, 57 were obtained from the
right hemisphere of DN, 103 from the left hemisphere of
CR, and 80 from the right hemisphere of CR. Of these 240
neurons, the discharge of 201 (84%) showed a significant
period effect, and the discharge of 228 (95%) showed a
significant direction effect based on the trial-by-trial ANOVA (P õ 0.01). All 240 neurons were task related, that
is their discharge exhibited either a significant period or
direction effect. The ANOVA is a highly sensitive test of a
directional effect (Crammond and Kalaska 1996) but does
not define the form of the directional modulation. Therefore
to define the type of directional modulation, the average
firing in the cue and track periods was fit to the cosine model
(Eq. 1). Significant directional tuning as defined by the
cosine model was found for 212/240 (88%) neurons. In
132, significant directional tuning was found for both the
cue and track periods. In 26 neurons, significant directional
tuning was present only during the cue period, and in another
54, the directional tuning was significant only in the track
period. The majority of the analysis focused on the relationship of the directional tuning in the cue and track periods
for cells in both PMd and MI.
01-26-99 09:03:21
neupal
LP-Neurophys
Downloaded from http://jn.physiology.org/ by 10.220.32.246 on July 31, 2017
FIG . 2. Tangential velocities are shown (left) for 4 tracking directions
(0, 90, 180, and 2707 ) and 4 target speeds (2, 3, 4, and 5 cm/s). Movement
trajectories for all 8 directions and 4 target speeds are also shown (right).
All traces are the averages of 10 trials. Dashed horizontal lines indicate the
tangential target velocity during the cue and track periods (i.e., ideal
tracking speed during the track period). All traces were aligned on the
initiation of the cue period, which is marked by dashed vertical lines. The
onset of the track period is denoted by solid vertical lines.
879
880
M.T.V. JOHNSON, J. D. COLTZ, M. C. HAGEN, AND T. J. EBNER
Based on the histological findings, the cells in this study
were recorded in PMd and MI. The locations of the penetrations corresponding to neurons with significant directional
tuning are shown in Fig. 5 for the three hemispheres. In the
left hemisphere of monkey CR (Fig. 5A), the recordings
were located primarily in PMd. The penetrations in the right
hemisphere of monkey CR (Fig. 5B) covered both PMd and
MI. In the right hemisphere of monkey DN (Fig. 5C), the
penetrations were predominantly in MI.
/ 9k2f$$de41
J278-8
An example of a neuron with significant directional tuning
for both the cue and track periods is shown in Fig. 6. The
discharge during the cue and track periods exhibited broad
directional tuning with increased firing for directions 2707
to 457. There was a reduction in firing below baseline at
1807 for both the cue and track periods. The fit to the cosine
tuning model in the cue period was highly significant over
the four target speeds, with the R 2 ranging from 0.94 to
0.99. The directional tuning during the track period was also
01-26-99 09:03:21
neupal
LP-Neurophys
Downloaded from http://jn.physiology.org/ by 10.220.32.246 on July 31, 2017
FIG . 3. Rectified electromyographic
(EMG) activity from 4 muscles during the
cue and track periods. Eight directions (0–
3157, in 457 increments) of the slowest 2cm/s target speed are shown. All traces result from averages of 10 trials aligned on
cue signal onset and are shown for each
muscle using the same scale. Dashed vertical lines denote the end of the hold period
and cue onset; solid vertical lines divide
the cue and track periods. Triangles above
the traces indicate when the EMG activity
differed by ú3 SD from the mean activity
during the hold period. Averaged epoch
EMG activity as a function of direction is
shown below the time course records.
VISUOMOTOR PROCESSING IN MOTOR CORTICAL NEURONS
881
over the four target speeds. The average absolute PD difference in the cue and track period was 1327 for this cell, and
the population data summarized in Fig. 10 shows that Ç50%
of the cells had an absolute PD difference of ú457. Again,
Downloaded from http://jn.physiology.org/ by 10.220.32.246 on July 31, 2017
FIG . 4. Horizontal (x) and vertical (y) components of the eye movements are shown over the temporal trial sequence and as trajectories over
the workspace. Four of the 8 directions of the tracking task (0, 90, 180,
and 2707 ) for the 2 extremes of target speeds, 2 cm/s (left) and 5 cm/s
(right), are shown. An additional 1,000-ms period of eye movement data
was included at the end of the tracking period. Leftmost dashed vertical
lines mark the end of the hold period; solid vertical lines divide the cue
and track periods. Rightmost dashed vertical lines mark the end of the track
period. Single trials of unsmoothed eye movements are shown.
highly significant with the R 2 ranging from 0.97 to 0.99.
The consistency of the firing on a trial-by-trial basis is demonstrated by the small standard deviations in the firing for
each direction. Furthermore, the tuning curves demonstrated
similar average PDs for the cue (358 { 67, mean { SD)
and track (355 { 47 ) periods over the four target speeds.
The depth of directional modulation was less for the cue
period than for the track period. The Idir averaged over the
four speeds was 0.50 { 0.06 for the cue period and 0.90 {
0.03 for the track period.
The PD of the discharge of a cell was not generally constant through the trial sequence, an example of which is
shown in Fig. 7. The discharge increased in the cue period
for directions 45–1807, an increase that persisted up to the
initial part of the track period. For directions ranging from
2257 to 07, there was considerably less activity above background in the cue period. The average PD during the cue
period was 92 { 127 (mean { SD) for the four speeds.
During the track period, the firing was tuned to the opposite
direction. The average PD of the track period was 320 { 47
/ 9k2f$$de41
J278-8
FIG . 5. Locations of the penetrations corresponding to neurons with
significant directional tuning for the 3 hemispheres studied. The left (A)
and right (B) hemispheres from monkey CR and the right hemisphere from
monkey DN (C) are shown. Electrode penetrations ( ● ) and marking lesions
( l ) are shown. The line drawn on the cortical surface is the estimate of
the dorsal premotor cortex and primary motor cortex boundary.
01-26-99 09:03:21
neupal
LP-Neurophys
882
M.T.V. JOHNSON, J. D. COLTZ, M. C. HAGEN, AND T. J. EBNER
the small standard deviations of the firing illustrate the consistency of the discharge over the individual trials. As for
the discharge of the neuron shown in Fig. 7, the depth of
modulation for the cue period was less than that for the track
period. The Idir for the cue period was 0.46 { 0.05 and for
the track period 0.66 { 0.04. For the subpopulation of 132
neurons with significant directional tuning in both the cue
and track periods, the Idir for the cue period, 0.34 { 0.21,
was significantly less than that for the track period, 0.67 {
0.31 (P õ 0.00001, n Å 132, paired Student’s t-test).
Neurons responding transiently to the onset of the cue target
have been found in the premotor cortex and have been referred
to as ‘‘signal neurons’’ (di Pellegrino and Wise 1993; Johnson
/ 9k2f$$de41
J278-8
et al. 1996b). These neurons, like that illustrated in Fig. 8, may
show directional tuning (di Pellegrino and Wise 1993; Johnson
et al. 1996b). For the cell shown in Fig. 8, firing transiently
increased for Ç500 ms after the onset of the cue target. This
signal-related firing was directionally tuned with a significant
cosine fit for three target speeds and had an average PD of
145 { 177. Directionality of the firing was less pronounced
relative to the ‘‘set neurons’’ shown in Figs. 6 and 7 as judged
by the smaller proportion of the variance accounted for by
the cosine model and the increased standard deviations of the
averaged data points. Overall, firing decreased in the track period for all directions and was not significantly directionally
tuned at any target speed. In the second half of the cue period
01-26-99 09:03:21
neupal
LP-Neurophys
Downloaded from http://jn.physiology.org/ by 10.220.32.246 on July 31, 2017
FIG . 6. Firing and speed profiles of a neuron with significant and congruent directional
tuning in both cue and track periods. Histograms of average firing rates binned in 20-ms
intervals are plotted over the trial sequence
of hold, cue, and track periods for the 8 target
directions and 4 target speeds (2, 3, 4, and 5
cm/s). Dashed vertical line denotes cue period onset, and solid vertical line denotes
track period onset. Note that at the higher
speeds the histograms appear less dense due
to the smaller number of 20-ms bins being
displayed. Averaged tangential hand velocity
is superimposed on the firing rate. Below each
set of histograms are plotted the cosine tuning
curves (
) of the averaged firing ( ● ) in
the cue (left) and track (right) periods. The
vertical bar above each averaged data point
indicates 1 SD for the firing from the 10 trials.
VISUOMOTOR PROCESSING IN MOTOR CORTICAL NEURONS
883
and the whole track period, the neuronal discharge was lower
than that during the initial baseline period. Of the 26 PMd
neurons with directional tuning only in the cue period, all demonstrated initial firing transients (i.e., signal-related). In the 132
neurons directionally tuned over both cue and track periods,
firing continued throughout the cue period. Thus the distinction
made by Weinrich and Wise (1982) between signal-related and
set-related activity has clear parallels in the present data.
Differences in directional tuning as a function of the task
sequence
Of the 132 neurons demonstrating significant directionality in both cue and track periods, many had PDs that differed
/ 9k2f$$de41
J278-8
between the cue and track periods (see Fig. 7). The distributions of the period PDs for the cue (Fig. 9A) and track (Fig.
9B) periods are shown in Fig. 9. Because both left and
right hemispheres were used for recording, the PDs were
normalized to a right hemisphere convention by mirror-reversing the PD of each left hemisphere neuron about the
vertical meridian (Funahashi et al. 1989; Georgopoulos et
al. 1982; Steinmetz et al. 1987). For the cue and track periods, the distributions of PDs were uniform about the workspace (Rayleigh test for uniformity, P ú 0.1 in both cases).
For a given neuron, the absolute difference in PDs (a range
from 0 to 1807 ) between the cue and track period was calculated (Fig. 9C). A predominance of vectors was in the 0 to
907 range, with a mean of 45.2 { 44.97. Almost one-half
01-26-99 09:03:21
neupal
LP-Neurophys
Downloaded from http://jn.physiology.org/ by 10.220.32.246 on July 31, 2017
FIG . 7. Neuron with significant but directionally noncongruent mean firing in the cue
and track periods. Discharge histograms with
superimposed tangential hand velocity traces
and cosine tuning curves use the same conventions as Fig. 6.
884
M.T.V. JOHNSON, J. D. COLTZ, M. C. HAGEN, AND T. J. EBNER
(47%) of the neurons had an absolute PD difference of
ú457. The PD distributions obtained from the cue and track
periods were significantly different (P õ 0.001, n Å 132,
Hotelling’s paired test) (Batschelet 1981). Furthermore, the
relative difference distribution (Fig. 9D) was significantly
different from zero along the x-axis (P õ 0.001), but not
along the y-axis (P ú 0.2, Student’s t-test on the sine and
cosine fractions with Bonferroni correction). This y-axis
symmetry implies that clockwise or counterclockwise deviations of the cue PD relative to the track PD were equally
probable. The relative difference plot (Fig. 9D) further demonstrates the equal probability of clockwise and counterclockwise differences between the cue and track PDs.
/ 9k2f$$de41
J278-8
Evaluation of the PD differences between cue and track
periods was performed at a higher temporal resolution by
dividing equally each period and recalculating the PDs for
each half period. The population distributions of the PDs
calculated from each of the four epochs (i.e., cue 1, cue 2,
track 1, and track 2) shown in Fig. 10 demonstrate uniform
directional tuning (P ú 0.1, Rayleigh test for uniformity).
For each neuron, absolute differences of the PDs were calculated for each period (cue 1, cue 2, track 1) relative to the
final tracking period (track 2). Relative to the late track
period, the mean change in PDs was greatest for the early
cue (79.27 ), less for the late cue (45.77 ), and least for the
early track (22.37 ) periods. The distributions differed sig-
01-26-99 09:03:21
neupal
LP-Neurophys
Downloaded from http://jn.physiology.org/ by 10.220.32.246 on July 31, 2017
FIG . 8. Neuron with increased firing
confined to the cue period. Discharge histograms with superimposed tangential velocity
traces and cosine tuning curves use the same
conventions as Figs. 6 and 7. A solid line is
drawn only through those data points with
significant directional tuning (R 2 ú 0.7).
VISUOMOTOR PROCESSING IN MOTOR CORTICAL NEURONS
nificantly (P õ 0.001 in all cases, Hotelling’s paired test).
The decreasing differences in the PD from the early cue to
the late track epochs suggests that a sequential, and not
abrupt, convergence of PDs is occurring. A similar analysis
comparing cue 1 and cue 2 to track 1 yielded a similar
systematic decrease in the mean differences in the PD as the
trial progressed. The mean differences in the PDs from cue
1 to track 1 and from cue 2 to track 1 were 42.1 and 23.5,
respectively. The difference between cue 1 and track 1 was
significant (P õ 0.001, Hotelling’s paired test).
Several other findings demonstrate that there is a gradient
of directional tuning properties across PMd and MI. The
prevalence of four properties of directional tuning were
mapped in 2-mm increments on the cortex from the central
sulcus (Fig. 11): 1) directional tuning occurring only during
the cue period, 2) directional tuning occurring only during
the track period, 3) directional tuning occurring in cue and
track that was õ457 divergent between the two periods, and
4) directional tuning in both periods that was ú457 divergent. As shown in Fig. 11A, neurons with significant directional tuning only in the track period comprised 45% of the
total at 2 mm from the central sulcus, and the percentage
decreased to 16% at the most anterior recording position
in the PMd. Conversely, the percentage of neurons with
significant directional tuning only in the cue period increased
from posterior to anterior with only 10% near the central
sulcus and 28% in the region of the PMd. As shown in Fig.
11B, the percentage of neurons with õ457 difference in cue
and track period PDs reached a maximum of 50% at 6 mm
from the central sulcus and decreased to Ç20% at 10 mm
from the central sulcus. Neurons with ú457 difference increased from 20% posteriorly to 37% anteriorly. It should
be noted that the percentage of total neurons with a cuetrack period PD difference of õ457 decreased near the central sulcus because the number of neurons with track only
directional tuning increased dramatically (see Fig. 11A).
The changes in these directional tuning properties as a function of distance were statistically significant (P õ 0.005, x 2
4 1 4 table). A 4 1 4 table (4 directional tuning properties
by 4 distance bins) was constructed by binning the data in
4-mm increments. A 4-mm binning was necessary so that
all expected cell counts were ú5, as required by the x 2
statistic (Devore and Peck 1993).
Temporal profiles of the preferred directions of individual
neurons
To evaluate the directional tuning of individual cells with
increased temporal resolution, the firing rate of these cells
was fitted to the cosine model (Eq. 1) in sequential 20-ms
bins. As described in METHODS, a significant fit (R 2 ú 0.7)
Functional-anatomic correlations
Of the 212 neurons with significant directional tuning in
either cue or track periods, 122 neurons were recorded in
PMd and 84 were in MI. Five penetrations resulting in six
directional neurons were located in the vicinity of the supplementary motor area (see Fig. 5A) and thus removed for the
comparisons of PMd and MI. Most of the neurons with significant cosine tuning limited to the cue period were located in
PMd (20 of a total of 26). For the 54 neurons with significant
directional tuning only in the track period, 26 were located
in PMd, whereas 28 were in MI. Neurons with directional
tuning in both cue and track periods (n Å 132) were also
found in both PMd (n Å 73) and MI (n Å 54). The proportions of these three types of directional neurons showed significant differences between the PMd and MI (P õ 0.001, x 2
3 1 2 table). Thus directional activity limited to the cue
period was more frequently found in PMd cortex, consistent
with previous observations (Alexander and Crutcher 1990;
Johnson et al. 1996b; Weinrich et al. 1984).
/ 9k2f$$de41
J278-8
FIG . 10. Distribution of the preferred directions, fractionated into early
and late cue (cue 1 and cue 2) and track (track 1 and track 2) periods, for
all neurons with significant directional tuning in 3 or 4 contiguous periods
including the final track 2 (top row). Absolute differences between cue 1,
cue 2, and track 1 relative to track 2 are shown (bottom row). The circular
means { SD are shown as a wedge superimposed on the absolute difference
distributions.
01-26-99 09:03:21
neupal
LP-Neurophys
Downloaded from http://jn.physiology.org/ by 10.220.32.246 on July 31, 2017
FIG . 9. Distribution of the preferred directions of all neurons with significant directional tuning (R 2 ú 0.7). Preferred directions were calculated
from averaged firing of the cue or track periods relative to the direction of
the moving target. A: distribution of the preferred direction (PD) vectors
obtained from the cue period shown with the magnitude equal to the R 2 .
B: distribution of the track period PD vectors with magnitude equal to R 2 .
The outer circle in A and B denotes an R 2 Å 1. C: distribution of the
absolute differences between cue and track PDs (range of distribution is
0–1807 ). D: distribution of the relative differences between the cue and
track PDs.
885
886
M.T.V. JOHNSON, J. D. COLTZ, M. C. HAGEN, AND T. J. EBNER
to the cosine model was required for each bin and a minimum
depth of modulation criterion within 1 SD of the average
Idir from the cue and track periods was imposed. Thus the
minimum Idir of a bin for the cue or track period was required
to be larger than 0.13 or 0.36, respectively. The resulting
significant PDs were used to map the temporal sequence of
the PDs, or PD trajectory through the trial. The PD trajectory
is referenced to the direction of the moving target for both
the cue and track periods (see METHODS ).
To define the temporal evolution of the PD, Fig. 12A
details the construction of a PD trajectory from the discharge
of one cell. Figure 12 also shows the PDs based on the
average firing for comparison. Tuning curves with the actual
data points for each consecutive 20-ms bin in the hold, cue,
and track periods are shown. The firing modulation during
the 1st 10 bins (200 ms) while the cursor was held in the
center hold period did not significantly fit the cosine tuning
model (dashed green lines). The fits of the 1st 10 of the
tuning curves (200 ms) in the cue period also did not meet
either the model fit (R 2 ú 0.7) or depth of modulation (Idir ú
0.13) criterion. Over the next 16 bins (320 ms), the fit of
the data met both criteria (solid red lines), and the PD
sequentially shifted from 90 to 07. Over the next 10 bins the
PD shifted further from 0 to 2707, after which the PD stabi-
/ 9k2f$$de41
J278-8
01-26-99 09:03:21
neupal
LP-Neurophys
Downloaded from http://jn.physiology.org/ by 10.220.32.246 on July 31, 2017
FIG . 11. Locations relative to the central sulcus of 4 functional types
of directional tuning. A: proportions of neurons with directional tuning only
in the cue period (Cue) and directional tuning only in the track period
(Track). B: proportions of neurons with PD differences between the cue
and track periods of õ457 and ú457. Although shown in 2 graphs, the
proportions of the 4 groups add to 100% for each 2-mm distance.
lized at 2707, and the depth of modulation progressively
increased over the next 14 bins until the start of the track
period. The maximum depth of modulation occurred 140 ms
(7 bins) before the onset of the track period. Through the
majority of the track period (solid blue lines), the PD remained constant at 2707, and the tuning curves gradually
decreased in modulation. During the last 100 ms of the track
period, there was a shift in the PD toward 2257. Note that
for 3 bins in the track period, the fit did not meet both criteria
(dashed blue lines).
Figure 12 demonstrates that there were smooth, gradual
shifts in the PD and in the depth of modulation over the
course of the trial. The correspondences between the peaks
in actual firing and the maxima of the tuning curves demonstrate that the model fits were not spurious. Furthermore, the
fits during the center hold period were invariably nonsignificant, as would be expected. The tuning curves and the firing
data, averaged over the hold, cue, and track periods are also
shown. The average discharge during the center hold period
was not significantly modulated. However, the directional
tuning curves obtained from the cue (R 2 Å 0.89, Idir Å 0.53,
PD Å 3127 ) and track periods (R 2 Å 0.82, Idir Å 0.85, PD Å
2787 ) were significant and fit the cosine tuning model. The
difference between the cue and track period PDs based on
the average firing in these two periods was 347, which minimizes the approximate 1707 shift (807 at the beginning of
the cue period to 2707 at the onset of track period) evident
in the PD trajectory that occurred over the cue period. The
calculation of the PD based on the average firing failed to
capture these temporal features of the directional tuning.
An additional analysis was undertaken to validate the PD
trajectories. The PD trajectory was calculated from averages
of odd and even trials taken from the randomized trial sequence for a single cell. The two PD trajectories as a function
of time resulting from this split averaging are shown in Fig.
12B for the same cell. In both sets of trials, the cue period
PD trajectory began around 807 and rotated clockwise to 07
and finally to 2707 to converge with the PD trajectory of
the track period. The tracking PD trajectory in both cases
continued along 2707 until the last third of the period. The
PD trajectory based on the entire set of trials is quite similar
(see Fig. 13B). Similar results were obtained for other cells.
Therefore several points establish the validity and reproducibility of the PD trajectory calculation based on 20-ms
bins. Previously, PD trajectories had been based on the calculation of the population vector and not the discharge of
individual cells (Georgopoulos et al. 1984, 1989a; Schwartz
1993). First, the criteria used to calculate the PD trajectory
were more stringent than typically used for epoch-based determinations of a cell’s PD (Caminiti et al. 1991; Fu et al.
1993; Georgopoulos et al. 1982; Schwartz 1992). In this
report we required not only statistical significance (R 2 ú
0.7) but also a minimum depth of modulation. Therefore the
directional tuning based on 20-ms bins is as rigorous and
valid (if not more so) than the directional tuning defined
over longer epochs. Second, as shown in Fig. 12, the individual tuning curves were nearly continuous and in complete
agreement with the PD calculations. Third, the determinations based on odd and even trials yielded similar results.
Over the 320 movement trials necessary to generate a complete data set in which the speed and direction of the trial
were randomized, the similarity of the results from the odd
VISUOMOTOR PROCESSING IN MOTOR CORTICAL NEURONS
887
and even trials reveals the reproducibility of the PD trajectory. Fourth, the fits based on the firing during the center
hold period were invariably nonsignificant, demonstrating
that the results from the cue and track period were not artifactual. Fifth, at least three reports from other investigators
using steplike reaching tasks have documented similar rotations in the PD trajectory when calculated on a finer time
scale (Baker et al. 1997; Mason et al. 1998; Scott and Kalaska 1996). Also, because these other reports were based
on a steplike reach, the observed rotations in the PD trajectory were not simply a function of the moving targets employed in this paradigm.
The PD trajectories ranged from those that were relatively
unidirectional over the cue and track periods to trajectories
that were nearly opposite in the cue relative to track periods.
Examples of PD trajectories from four neurons are shown
/ 9k2f$$de41
J278-8
in Fig. 13. In this figure, the neurons are presented in terms
of increasing absolute difference between the cue and track
period PDs, with differences of 33, 34, 47, and 1237 in A,
B, C, and D, respectively. The PD trajectories reveal that
the similarities and differences in the directional tuning during the cue and track periods are much more complex than
revealed by the period average PDs. For the example shown
in Fig. 13A, the PD trajectory became significant (R 2 ú
0.7) Ç300 ms after the onset of the cue period. The PD was
initially 907 but shifted to 1607 by the end of the cue period,
rotating counter clockwise with an average velocity of 1007 /
s. During the track period, the PD trajectory initially rotated
another 207 counterclockwise before maintaining a constant
direction of Ç1807 for the remaining two-thirds of the period. For the last 150 ms of the track period, the PD calculations were not significant. Examination of the firing profile
01-26-99 09:03:21
neupal
LP-Neurophys
Downloaded from http://jn.physiology.org/ by 10.220.32.246 on July 31, 2017
FIG . 12. Sequence of tuning curves derived from sequential 20-ms bins are shown
in A for the hold (green), cue (red), and
track (blue) periods. All records are averages of 10 trials for the fastest 5-cm/s target
speed. Solid sinusoidal tuning curves are
shown for significant fits (R 2 ú 0.7 and
Idir ú period Idir 0 1 SD), whereas dotted
curves signify fits not satisfying both criteria. Data points from which the tuning
curves were derived are superimposed.
Shown below the sequence of binned tuning curves are the tuning curves and firing
data averaged over the hold, cue, and track
periods. The firing rate calibration bar applies to both binned and period tuning
curves. The x-axis denotes the PD convention based on the direction of the moving
target. In B, PD trajectories as a function
of time are shown for averages based on the
even (blue symbols) and odd trials (purple
symbols) for the same neuron. The y-axis
is time (the cue to track period sequence
lasts 2 s for the 5-cm/s target speed), and
the x-axis is the PD (ranging from 0 to
3607 ). The PD trajectories were referenced
to the direction of the moving target (see
METHODS ). Therefore for this cell the PD
trajectory pointed to 907 initially in the cue
period and rotated clockwise to 2707 at the
start of the track period.
888
M.T.V. JOHNSON, J. D. COLTZ, M. C. HAGEN, AND T. J. EBNER
reveals a similar picture. Firing relative to the hold period
increased Ç200 ms after the cue period onset but did not
become directionally tuned for another 100 ms. The PD
trajectory followed the ridge of increased firing (orangered pseudocolor) with remarkable fidelity. The confidence
interval was initially {707, reflecting the smaller depth of
modulation in the early cue period. During the cue period
the confidence interval constricted to {307 and decreased to
approximately {107 midway through the track period. The
confidence interval confirms that the changes in PD in the
cue period exceed the uncertainty of the estimate.
In examples B, C, and D, the PD trajectories in the cue
period demonstrated initial rotations covering up to 1807 of
the workspace and then aligned during the cue period to
approximate the direction of the track PD trajectories. These
examples have several common features. First, the direction
modulation of the discharge was delayed relative to the appearance of the target by 150–200 ms. Baseline firing values
evolved into significant directional tuning, and, subsequently, a PD trajectory developed. Second, the rotations of
the PD trajectories were most marked in the cue period. The
PD trajectory rotated in the clockwise direction in two of
the cells (B and D) and counterclockwise in two (A and
C). Third, the PD trajectories during the track period were
relatively straight and approximated the track period average
PD. In B the PD trajectory was initially directed to 907 and
rotated at Ç2007 /s to 2707 by the end of the cue period.
The 1807 shift of the PD trajectory in the cue period was
larger than the 5–95% confidence limit. In C the PD trajectory started at 1357, shifting to 3007 after a rotation of Ç2007 /
s. Once the track period began, the trajectory remained
within {457 of the period average PD of 3017. In D the PD
trajectory shifted clockwise from 45 to 3007 with a rotational
velocity of 2107 /s. The PD trajectory continued to shift
clockwise another 707 until the initial third of the track period, stabilizing around the period average PD of 2267. The
/ 9k2f$$de41
J278-8
confidence intervals demonstrate that the shifts during the
cue period were statistically significant.
Temporal characteristics of the PD trajectories
From the slowest to fastest target speeds, the available
time allowed for the transformation of directional visual information into directional motor commands ranged from 2.5
to 1 s. Because temporal constraints were found in the oculomotor behavior, the temporal profiles of the PD trajectories
were also analyzed for similar constraints. The average latency of the directional discharge ranged from 299 to 388
ms for the fastest to slowest target speeds, respectively (Table 1). Linear regression analysis demonstrated that the average onset latencies significantly increased with decreasing
target speed, with a slope of 033.3 ms per cm/s (F-ratio Å
11.5, P õ 0.001). Thus the detection of target directionality
was delayed in the longer cue periods resulting from slower
target speeds. Conversely, when the cue period was limited
to 1 s by the 5-cm/s target, the onset of directional tuning
occurred earlier. Directional tuning was maintained for a
larger fraction of the trial duration as the target speed increased, resulting in a significant linear regression relationship (slope Å 3.6% trial duration per cm/s, F-ratio Å 15.3,
P õ 0.0001). Thus, when the cue period was limited by
faster target speeds, directional tuning occurred 20% earlier
and occupied 20% more of the time allotted before movement. These target speed-related changes in the onset and
duration of directional tuning paralleled the changes seen in
oculomotor behavior. As the cue period became longer,
smooth pursuit was present for a smaller fraction of the cue
period (Fig. 4).
Rotations of the PD trajectories were prominent, particularly in the cue period (see Figs. 12 and 13). Group averages
of the maximal rotational velocity ranged from 438 to 5527 /
s and from 297 to 3237 /s for the cue and track periods,
01-26-99 09:03:21
neupal
LP-Neurophys
Downloaded from http://jn.physiology.org/ by 10.220.32.246 on July 31, 2017
FIG . 13. Contour plots of the firing and
PD trajectories for 4 neurons. The neurons
are ordered in increasing difference between cue and track epoch based PDs
(ranging from 33 to 1237 ). The x-axis is
the trial sequence from cue to track (a total
duration of 2 s, corresponding to a 10-cm
target travel). The PD is mapped on the yaxis, ranging from 0 to 3607. The amplitude
of firing relative to the hold period is coded
by color (peak orange, minima purple).
Significant preferred directions (by R 2 and
Idir criteria) calculated in 20-ms bins are
marked by solid symbols. Confidence intervals (5–95%, based on the von Mises distribution) for the location of the PD are
marked by smaller symbols on either side
of the PD trajectory. All records are averages of 10 trials for the fastest 5-cm/s target
speed.
VISUOMOTOR PROCESSING IN MOTOR CORTICAL NEURONS
TABLE
1.
889
PD trajectory onset latency, duration, and rotational velocity as a function of target speed
Instantaneous PD Angular Speed, deg/s
Cue Period
Target Speed
2
3
4
5
cm/s
cm/s
cm/s
cm/s
Latency*
388
385
316
299
{
{
{
{
259
235
182
168
Duration†
42
48
54
52
{
{
{
{
20
22
22
22
Mean
155
155
156
160
{
{
{
{
85
88
110
85
Track Period
Maximum
552
451
440
438
{
{
{
{
248
216
217
212
Mean
104
103
100
110
{
{
{
{
50
51
47
57
Maximum
323
313
293
297
{
{
{
{
130
145
203
137
Values are means { SD; number of neurons is 132. PD, preferred direction. * Onset of directionality in ms. † Percent of trial that is directionally
significant.
Composite profiles of the PD trajectories
A composite of PD trajectories is shown for the extremes
of target speed in Fig. 14A to illustrate the temporal evolution of directional tuning. These trajectories were obtained
from 20 neurons selected from the sampled population on
the basis of having the largest number of contiguously significant PD trajectories. The selected sample had a uniform
distribution of PDs in both cue and track periods (Rayleigh
test for uniformity, P ú 0.3 in both cases) and was characteristic of the entire population of cells (Fig. 14B). The track
period PD was first rotated to align on 07, and the PD trajectories for the cue and track periods were rotated by an equal
amount.
This population of trajectories confirms that the PD differed during the cue and track periods. During the initial cue
period, the PD could differ by as much as 1807 from the
track PD. For both 5- and 2-cm/s target speeds, the PD
trajectory reached initial significance 100–200 ms after onset of the cue target, corresponding to the initiation of oculomotor smooth pursuit (Fig. 4). At either extreme of target
speed, the PD trajectories during the cue period characteristically rotated over one to two quadrants of the workspace
until the cue-track transition. Around this point, the PD trajectory converged to that of the track period PD (normalized
to 07 ), and the PD trajectories during the track period were
relatively straight.
For the fastest target speed (Fig. 14A, top), the PD trajectories converged at the onset of the track period to a zone
of {307 of the normalized track period PD. For the slowest
target speed (Fig. 14A, bottom), the PD trajectories did not
fully converge to within {307 of the normalized average
until 500 ms after the track period onset. It follows from the
/ 9k2f$$de41
J278-8
design of the error constraint of the paradigm and the observed kinematics that the monkey did not have to move at
the exact start of the track period for the slower target speeds.
Movements could be delayed as much as 600 ms, corresponding to an error window of 1.2 cm for a target speed
of 2 cm/s. The kinematic records revealed that movement
was commonly delayed, and the hand did not reach target
speed until 500 ms after the onset of the track period (see
Figs. 2 and 5–7). Thus convergence of the PD trajectories
to the preferred direction of the track period occurred in the
time period in which the movement was being initiated. It
has been shown that the discharge of premotor neurons is
modulated by gaze angle (Boussaoud 1995). However, it
should be stressed that the angular rotation of the cue period
PD trajectories was not simply due to the eye movements
because the saccadic eye movements in the earlier part of the
cue covered the entire workspace (see Fig. 4). Furthermore,
saccadic behavior was qualitatively similar during both the
cue and track periods, yet the rotation of the PD trajectories
occurred primarily in the cue period.
If the PD trajectory behavior reflects only a directional
visuomotor process, it should be invariant with respect to
target speed. The directional tuning as determined from period averages of firing did not change systematically with
target speed. For each neuron, the mean standard deviation
of the PD over the four target speeds was 18.8 (range 2.8–
58.9) and 11.0 (range 1.4–39.9) for the cue and track period
averages, respectively. The PD trajectories were also similar
across the four target speeds. The PD trajectories of Fig. 14
were compared bin by bin for the extremes of target speed.
The 2-cm/s speed trials were reduced to 100 bins, the same
size as the 5-cm/s trials. Bins with significant directionality
from the 2- and 5-cm/s trials were plotted as x-y pairs in
Fig. 14C, and a linear regression between the pairs was
computed. The R 2 was 0.80, showing that at the extremes
of the tracking speeds the PD trajectories were highly correlated. Thus, whether calculated over periods or every 20 ms,
the preferred direction varied minimally with target speed.
The means { SD of the PD trajectories were calculated
using all 132 neurons that demonstrated significant directional tuning in both the cue and track periods (Fig. 15).
The average track PD was normalized to 07 using the same
conventions as Fig. 14. The convergence of the PD trajectories shown in Fig. 14 parallels the constriction of the variability shown in Fig. 15. For the 5-cm/s tracking speed (top
graph) the mean was approximately zero through the trial
sequence as approximately similar proportions of PD trajectories rotated clockwise and counterclockwise relative to the
01-26-99 09:03:21
neupal
LP-Neurophys
Downloaded from http://jn.physiology.org/ by 10.220.32.246 on July 31, 2017
respectively (Table 1). The difference in PD trajectory rotational velocity between the cue and track periods was highly
significant (P õ 0.001, at all target speeds, paired Student’s
t-test), whereas the differences due to target speed were not.
Linear regression of the mean rotational velocity to target
speed did not result in a significant relationship for the cue
or track periods (P ú 0.4, in both cases). Thus the rotational
velocity of the PD trajectory was a function of the visuomotor processing occurring in the cue relative to the track periods but not the target speed, further supporting the conclusion that the PD trajectory was relatively invariant with respect to speed. Furthermore, a fixed rotational velocity would
explain the increase in the relative duration of the PD trajectory during the shorter cue periods.
890
M.T.V. JOHNSON, J. D. COLTZ, M. C. HAGEN, AND T. J. EBNER
over the cue-track transition followed the same pattern for
the extremes of target speed. If the alignment of the PD
trajectory with the eventual track PD reflects a visuomotor
process, this process occurred at the cue-track transition.
The convergence of the PD trajectories (Fig. 14) and the
constriction of the variance (Fig. 15) both were delayed at
the slower speed, arguing for a role in movement initiation.
DISCUSSION
Directional coding during a visuomotor task
FIG . 14. Temporal profiles of contiguously significant PD trajectories
from 20 neurons. PD as a function of time throughout the trial sequence is
shown in A for the fastest 5-cm/s and slowest 2-cm/s target speeds. Vertical
line at time 0 denotes onset of the track period. The PD trajectories were
rotated such that the track period PD trajectory was aligned with 07. The
distributions of the average PDs from the cue and track periods are shown
in B using the convention of Figs. 9 and 10. Comparison of the PD trajectories from the 2- and 5-cm/s target speed trials is shown in C. The 2-cm/s
speed trials were reduced to 100 bins, the same size as the 5-cm/s trials.
Bins with significant directional tuning from the 2- and 5-cm/s trials were
plotted as x-y pairs. The linear regression line and the 99% confidence
intervals are shown. The R 2 was 0.80. Axes are normalized as in A to 0 {
1807.
final track period PD. The SD envelope constricted throughout the 1-s cue period, ranging initially from {1007 at the
onset of the cue period to {107 at 300 ms into the track
period. For the 2-cm/s target speed trials (bottom graph),
the mean PD was also approximately zero. The SD was
initially very large 300 ms after the cue onset and then
decreased to approximately {907 and remained at that level
through the next 1.2 s of the cue period. The SD then began
to decrease Ç1 s before track onset, constricting to its smallest value of about {107 at 400 ms into the track period.
Thus the evolution of the variance in the PD trajectories
/ 9k2f$$de41
J278-8
FIG . 15. Temporal profiles of the means { SD for the PD trajectories
of all 132 neurons with significant directional tuning in both the cue and
track periods. The mean (central solid line) and {SD (dotted lines) of the
PD trajectories as a function of time through the trial sequence are shown
for the fastest 5-cm/s (top) and slowest 2-cm/s (bottom) target speeds.
Vertical line at time 0 denotes onset of the track period. PD trajectories
were normalized by setting the average track period PD to 07, the same
convention used in Fig. 14. Thus the y-axes range from 0 { 1807.
01-26-99 09:03:21
neupal
LP-Neurophys
Downloaded from http://jn.physiology.org/ by 10.220.32.246 on July 31, 2017
Three main findings emerged concerning the discharge of
primary motor and premotor neurons during this instructed
delay, pursuit tracking task. Both the instructed delay (cue)
and the movement (track) periods differed from previously
used center-out paradigms based on step movements. This
task required visually guided, error-constrained arm tracking
of a constant speed target, as opposed to fast steps of unconstrained trajectory and speed to a stationary target (Fu et al.
1993; Georgopoulos et al. 1982; Kalaska et al. 1983, 1989;
Wise et al. 1992). Step movements first require detection
of the onset and position of the target before a largely feedforward and stereotyped movement is made. The movement
is characterized by a bell-shaped tangential velocity profile,
and there is little reliance on visual feedback until termination of the movement (see Georgopoulos 1986). In contrast,
pursuit tracking requires continuous correction of error, re-
VISUOMOTOR PROCESSING IN MOTOR CORTICAL NEURONS
/ 9k2f$$de41
J278-8
cue and track periods were more common in PMd, and the
difference decreased from PMd to MI. These findings are
consistent with the hypothesized role for PMd in the mapping of complex visuomotor relationships (Boussaoud and
Wise 1993; di Pellegrino and Wise 1993).
PD trajectory and coordinate systems
The observed rotation in the PD trajectories may be a
function of the coordinate system selected for the calculation
of the PD, i.e., the workspace-centered system used in this
report. It is acknowledged that the PD trajectories may differ
in another, unidentified coordinate system. For example, it
is not likely that a single system such as a joint or muscle
based coordinate system can explain the rotations in the PD
trajectories. First, the largest rotations in the PD trajectories
occurred in the early cue period when there was no movement. It is highly unlikely that the discharge during the
cue period is encoded in joint space when there is no joint
movement, particularly given the extended duration of the
cue period (1–2.5 s). Neurons in the PMd and MI are modulated by visuospatial aspects of moving to a target (di Pellegrino and Wise 1993; Shen and Alexander 1997a,b; Zhang
et al. 1997), and it is unlikely that these neurons encode
this visuospatial information in joint space or muscle coordinates. Second, the PD trajectories were straightest in the
track period when rotation of the joints and contraction of
the muscles were occurring. One would anticipate that the
PD trajectories would show the greatest rotations just before
or in the track period if the neuronal discharge was encoded
in a joint space coordinate frame.
However, the findings are consistent with the concept that
the visuomotor transformation may involve a change in coordinate systems in time (Kalaska and Crammond 1992; Shen
and Alexander 1997a; Soechting and Flanders 1992). Recent
evidence from single-unit recordings in the PMd and MI
suggests that visuospatial information is processed first, possibly in an extrinsic coordinate frame, and that information
about the actual limb movement is processed later, possibly
in an intrinsic frame (di Pellegrino and Wise 1993; Shen
and Alexander 1997a,b). The rotation of the PD in the cue
period could reflect a gradual change in the coordinate
frames in which the information is represented. The present
findings are consistent with single neurons participating in
such a process (Zhang et al. 1997), in addition to the participation of different neuronal populations (Lurito et al. 1991;
Shen and Alexander 1997a,b).
PD trajectory and multiplexed signal processing
Changes in the PD trajectory could be the result of multiplexing of other sensory or motor parameters within the
neuronal discharge at specific times during the task. For
example, changes in starting position or arm posture can
alter the directional tuning of MI neurons (Caminiti et al.
1991; Scott and Kalaska 1995, 1997). In this task both target
detection and speed identification are likely candidates, signaling the onset of the task and the onset of movement.
Evidence for detection is the finding of transient neuronal
discharge shortly after the visual cue presentation. This type
of ‘‘signal’’ activity is well documented in the PMd (Crammond and Kalaska 1994; Johnson et al. 1996b; Wise et al.
01-26-99 09:03:21
neupal
LP-Neurophys
Downloaded from http://jn.physiology.org/ by 10.220.32.246 on July 31, 2017
sulting in a tangential velocity profile approximating that of
the target. The duration of the tracking movements in this
task was 1–2.5 s as opposed to 200–400 ms movement
times typical of step movements over roughly equivalent
distances (Fu et al. 1993; Georgopoulos et al. 1982; Kalaska
et al. 1989; Ojakangas and Ebner 1992). Therefore in this
task the observation time for neural and behavioral events
was increased by a factor of 2–5. Because the directional
tuning of the discharge occurred throughout the entire cue
period for up to 2.5 s, more detailed observations on the
temporal processing of visual information were possible. In
addition, the pursuit tracking task allowed independent control of movement direction and speed. It should also be
stressed that this task differs from ‘‘nonpursuit’’ tracking
paradigms in which the animal is required to duplicate successively presented trajectory templates without a speed constraint (Schwartz 1992, 1993).
First, directional tuning was a prominent feature of both
the cue and track periods, demonstrating the robustness of
the directional information present in a variety of tasks
(Caminiti et al. 1991; Fu et al. 1993, 1995; Georgopoulos
et al. 1982; Johnson et al. 1996b; Kalaska et al. 1983; Shen
and Alexander 1997a,b). Second, a cell’s PD during the cue
period did not always point in the same direction of its PD
during the track period. Over all cells, the mean cue period
PDs differed significantly from those of the track period
PDs, and approximately half of the sampled neurons had
differences ú457 between the PD of the cue and track periods. These differences in period PDs became more apparent
when the cue and track periods were halved. The PDs obtained from the early cue period demonstrated a greater divergence relative to the tracking PDs than did the PDs from
the later cue period. Thus a cell’s PD was not constant across
the trial. The third and somewhat unexpected finding was
that sequential PDs calculated in 20-ms bins rotated from
an initial direction in the cue period to align with the PD
trajectory of the track period. During most of the track period, the PD trajectory was relatively constant. The global
features of the PD trajectory, including its rotational velocity,
varied little over different target speeds. Similar rotation in
the PD trajectories of PMd neurons has been observed in an
instructed delay reaching task (Mason et al. 1998). Furthermore, the detailed structure of the PD trajectory differed as
a function of the trial sequence, revealing larger rotations and
significantly higher rotational velocities in the cue relative to
the track period. In the next section the implications of the
PD trajectories in terms of coordinate systems, multiplexing
of signals, visuomotor transformations, and population vector rotations are discussed.
Last, the results confirm and extend previous findings of
a spatial gradient reflecting visuomotor processing across
PMd and MI. Premovement neural activity has been shown
to decrease rostral-caudally from PMd to MI (Alexander
and Crutcher 1990; Johnson et al. 1996b; Weinrich et al.
1984). A similar gradient in which neurons encoding visuospatial aspects of movements are preferentially located in
PMd and neurons encoding movement in MI has been described (Shen and Alexander 1997a,b). In this study the
visuomotor properties of directional tuning exhibited a rostral-caudal gradient across PMd and MI. Directional tuning
in the cue period was more prevalent in PMd than MI. Neurons with the greatest differences in their PDs between the
891
892
M.T.V. JOHNSON, J. D. COLTZ, M. C. HAGEN, AND T. J. EBNER
PD trajectory and visuomotor transformations
Initially in the cue period, all directions of the workspace
are covered by a population of PD trajectories, an optimal
situation for target detection (see Figs. 14 and 15). The
shifts in the PD during the cue period may provide a wide
aperture for single cells to detect targets moving at various
directions. Once detected, the direction of the target and the
direction of the upcoming movement can be processed.
Along with direction, the speed of the target must also be
identified. The onset of movement in this paradigm was not
dictated by an explicit ‘‘GO’’ signal, rather it was governed
by the speed of the target during the cue period. Preliminary
analysis showed that the firing in the cue period was modulated by the speed of the target (Johnson et al. 1997). The
directional tuning of the discharge gradually constricted to
align with the PD of the actual movement. This can be
appreciated in the plot of the individual PD trajectories (Fig.
14) and in the plot of the standard deviation of the PD
trajectories (Fig. 15). In either case, the directional tuning
in the cue aligned within {107 of the track PD over the cuetrack transition. Over the extremes of target speeds, this
convergence occurred from a period extending from 1 s
before track onset until Ç400 ms into the track period. Thus
the convergence of the PD trajectories may provide an intrinsic ‘‘GO’’ signal. Another factor that may have contributed
to the generation of the movement was the increased Idir
found for the track period.
One feature of the rotations in the PD is that for the slower
tracking speeds the shift in the PD took longer than for the
/ 9k2f$$de41
J278-8
higher speeds (Table 1, Figs. 14 and 15). Despite the fact
that for the slower tracking speeds the onset for directional
tuning was delayed and the proportion of significant directional tuning decreased in the cue period (Table 1), still the
slower speeds were characterized by a longer PD trajectory.
Why should the process take longer for the slower tracking
speeds? We suggest that this finding is consistent with the
hypothesis that the convergence of the PD trajectories to the
track preferred direction is also timing signal for the onset
of hand tracking. Irrespective of the duration of the cue
period, the convergence of PD trajectories was completed
at the cue-track transition. The convergence could not occur
earlier if it was the signal to move. It could be argued that
the visuomotor transformation is completed in a much
shorter period of time. In instructed-delay reaching tasks,
both PMd and MI neurons can show sustained discharge in
relation to the cue. However, this set-related activity is not
necessarily directionally constant. In fact, there are many
examples in the literature in which the discharge during the
instructed delay period gradually changes even though the
cue location is fixed [ for examples, see Fig. 4 in Wise and
Mauritz (1985), Figs. 2 and 12 in Weinrich et al. (1984),
and Figs. 1–4 in Shen and Alexander (1997b)]. The rotations in the PDs described in this report are completely consistent with these gradual directional changes in set-related
activity.
After the detection of the target, the visual stimulus must
be transformed into a motor command. Pertinent to the present results is the question of how the direction of the moving
cue target is transformed into the upcoming movement direction. Three lines of evidence argue against a direct spatial
mapping of cue target direction relative to the eventual
movement direction. First, the cue period PD could point in
any of the eight target directions irrespective of the track
period PD (see Fig. 14). Second, the absolute PD differences
between the cue and track periods spanned a continuum
between 0 and 1807 (see Figs. 8 and 9). If a strict directional
mapping of the cue target position was occurring, the fixed
and linear movement of the target would not be expected to
produce shifts in the PD across the trial. Third, the systematic
rotation and convergence of the PD trajectory during the
cue period suggests that the directional information obtained
from the cue is not simply mapped onto the directional discharge during the track period; rather, this suggests that a
more complex sensorimotor transformation must be occurring (Wise et al. 1996).
More complex sensorimotor transforms include nonstandard stimulus response mapping, in which stimuli with limited or incongruous spatial information must be mapped onto
the appropriate motor behavior (Crammond and Kalaska
1994; Fitts and Seeger 1953; Kornblum et al. 1990; Wise et
al. 1996). One of the proposed roles of the PMd is the
learning and computation of nonstandard sensorimotor transformations (Wise et al. 1996). A similar hypothesis has
been advanced for the parietal cortex (Crammond and Kalaska 1994). Premovement firing in instructed delay tasks
has been proposed to reflect the selection of an appropriate
response relative to a stimulus (Crammond and Kalaska
1994; Wise et al. 1996). The prevalence of divergent cue
and track PDs was greater for the PMd than for MI neurons.
Consistent with the posited role for PMd in sensorimotor
transformations, the differences in the PD in the cue and
01-26-99 09:03:21
neupal
LP-Neurophys
Downloaded from http://jn.physiology.org/ by 10.220.32.246 on July 31, 2017
1992), and several cells demonstrated this signal characteristic in this pursuit tracking paradigm. However, the transient
nature of this firing is unlikely to account for the prolonged
rotations in the PD trajectory. Preliminary evidence based on
multiple regression analyses of speed and direction coding in
these neurons suggests that speed is encoded jointly with
direction during the cue period (Johnson et al. 1997). Therefore it is possible that the encoding of other parameters
altered the directional tuning.
Changes in the PD trajectory could also be the result of
the multiplexing of other control functions, including eye
movements (Boussaoud 1995) and selective attention (di
Pellegrino and Wise 1993). Eye movements followed a sequence through the trial of 1) initial randomly directed saccades, 2) saccades to target, 3) smooth pursuit of the cue
target intermittently broken by randomly directed saccades,
4) smooth pursuit of the target during the initial half of the
track period, and 5) saccades during the end of the track
period. The eye movements were symmetrical in the cue
and track periods. Selective attention would be expected to
follow a similar symmetrical time course, initially directed
over the entire workspace, then directed at the moving target,
and finally drifting off at the end of the track period. However, the PD trajectories were not symmetrical, exhibiting
substantial shifts in the PD during the cue period while the
PD was relatively constant in the track period. Therefore
oculomotor behavior and/or selective attention are unlikely
to have contributed to the changes in the PD trajectory during
the initial cue period. However, irrespective of the contributing factors, the net effect is to modify the directional tuning
of these cells.
VISUOMOTOR PROCESSING IN MOTOR CORTICAL NEURONS
Implications for population vector rotation
The rotation of the PD during the long cue period in this
paradigm may provide a mechanism for the rotation of the
population vector that is initially tuned toward a visual target
and eventually redirects toward the upcoming movement
(Georgopoulos et al. 1989b; Lurito et al. 1991). For an
instructed redirection task, angular changes in the PD have
been noted at a population level but have not been described
for individual neurons (Lurito et al. 1991). The rotation of
the PD trajectory in individual cells can be implicated as a
mechanism underlying the rotation of the population vector.
Similar rotational velocities for both the PD trajectory and
population vector support this idea. Rotational velocities of
732 and 4197 /s have been found for the rotations of the
population vector for a 907 visuomotor transformation
(Georgopoulos et al. 1989b; Lurito et al. 1991). Peak rotational velocity averaged over all cells was Ç4507 /s, with
individual neurons having peak velocities as high as 1,1007 /
s (Table 1). The rotation of the PD trajectory in individual
cells could also account for the relationship between the
reaction time and the degree of rotation observed in psychophysical studies (Georgopoulos et al. 1989b; Lurito et al.
1991; Shepard and Cooper 1982). This monotonic relationship of reaction time and amount of rotation requires that
the ‘‘mental’’ rotational velocity remain constant. The PD
trajectory rotation fits this criterion, because it is constant
relative to target speed and over marked differences in cue
duration. Thus the rotations of PD trajectory in individual
neurons may contribute to the rotation of the population
vector (Georgopoulos et al. 1989b; Lurito et al. 1991).
We thank M. McPhee for assistance with graphics and histology and J.
Bailey and L. King for help with manuscript preparation. E. Ebner assisted
in programming. Dr. Christopher Bingham was consulted on statistical
methodology.
This work was supported by National Institutes of Health Grants 5R01NS-31530 and a grant from the Human Frontier Science Program. M.T.V.
Johnson was supported by T32-NS-07361 and J. D. Coltz by F31-MH11430.
Address for reprint requests: T. J. Ebner, Dept. of Neurosurgery, Univer-
/ 9k2f$$de41
J278-8
sity of Minnesota, Lions Research Building, 2001 Sixth St. SE, Minneapolis,
MN 55455.
Received 10 April 1998; accepted in final form 31 August 1998.
REFERENCES
ALBRIGHT, T. D. Direction and orientation selectivity of neurons in visual
area MT of the macaque. J. Neurophysiol. 52: 1106–1130, 1984.
ALEXANDER, G. E. AND CRUTCHER, M. D. Preparation for movement: neural
representations of intended direction in three motor areas of the monkey.
J. Neurophysiol. 64: 133–150, 1990.
ASHE, J. Force and the motor cortex. Behav. Brain Res. 86: 1–15, 1997.
BAKER, J. T., FELLOWS, M., AND DONOGHUE, J. P. Directional tuning rotations in motor cortical neurons during radial arm movements. Soc. Neurosci. Abstr. 23: 1555, 1997.
BATSCHELET, E. Circular Statistics in Biology. New York: Academic, 1981.
BOUSSAOUD, D. Primate premotor cortex: modulation of preparatory neuronal activity by gaze angle. J. Neurophysiol. 73: 886–890, 1995.
BOUSSAOUD, D. AND WISE, S. P. Primate frontal cortex: neuronal activity
following attentional versus intentional cues. Exp. Brain Res. 95: 15–
27, 1993.
CAMINITI, R., JOHNSON, P. B., GALLI, C., FERRAINA, S., AND BURNOD, Y.
Making arm movements within different parts of space: the premotor
and motor cortical representation of a coordinate system for reaching to
visual targets. J. Neurosci. 11: 1182–1197, 1991.
CRAMMOND, D. J. AND KALASK A, J. F. Modulation of preparatory neural
activity in dorsal premotor cortex due to stimulus-response compatibility.
J. Neurophysiol. 71: 1281–1284, 1994.
CRAMMOND, D. J. AND KALASK A, J. F. Differential relation of discharge in
primary motor cortex and premotor cortex to movements versus actively
maintained postures during a reaching task. Exp. Brain Res. 108: 45–
61, 1996.
CRUTCHER, M. D. AND ALEXANDER, G. E. Movement-related neuronal activity selectively coding either direction or muscle pattern in three motor
areas of the monkey. J. Neurophysiol. 64: 151–163, 1990.
COLBY, C. L. AND DUHAMEL, J. Heterogeneity of extrastriate visual areas
and multiple parietal areas in the macaque monkey. Neuropsychologia
29: 517–537, 1991.
DEVORE, J. AND PECK, R. Statistics (2nd ed.). Belmont, CA: Wadsworth,
1993.
DI PELLEGRINO, G. AND WISE, S. P. Visuospatial versus visuomotor activity
in the premotor and prefrontal cortex of a primate. J. Neurosci. 13: 1227–
1243, 1993.
DUM, R. AND STRICK, P. L. The origin of corticospinal projections from
the premotor areas in the frontal lobe. J. Neurosci. 11: 667–689, 1991.
FITTS, P. M. AND SEEGER, C. M. S-R compatibility: spatial characteristics
of stimulus and response codes. J. Exp. Psychol. 46: 199–210, 1953.
FLAMENT, D., ONSTOTT, D., FU, Q.-G., AND EBNER, T. Distance- and errorrelated discharge of cell in premotor cortex of rhesus monkeys. Neurosci.
Lett. 153: 144–148, 1993.
FU, Q.-G., FLAMENT, D., COLTZ, J. D., AND EBNER, T. J. Temporal encoding
of movement kinematics in the discharge of primate primary motor and
premotor neurons. J. Neurophysiol. 73: 836–854, 1995.
FU, Q.-G., SUAREZ, J. I., AND EBNER, T. J. Neuronal specification of direction and distance during reaching movements in the superior precentral
premotor area and primary motor cortex of monkeys. J. Neurophysiol.
70: 2097–2116, 1993.
FUNAHASHI, S., BRUCE, C. J., AND GOLDMAN-RAKIC, P. S. Mnemonic coding of visual space in the monkey’s dorsolateral prefrontal cortex. J.
Neurophysiol. 61: 331–349, 1989.
GEORGOPOULOS, A. P. On reaching. Annu. Rev. Neurosci. 9: 147–170, 1986.
GEORGOPOULOS, A. P., CRUTCHER, M. D., AND SCHWARTZ, A. B. Cognitive
spatial-motor processes 3. Motor cortical prediction of movement direction during an instructed delay period. Exp. Brain Res. 75: 183–194,
1989a.
GEORGOPOULOS, A. P., KALASK A, J., CAMINITI, R., AND MASSEY, J. T. On
the relations between the direction of two-dimensional arm movements
and cell discharge in primate motor cortex. J. Neurosci. 2: 1527–1537,
1982.
GEORGOPOULOS, A. P., KALASK A, J. F., CRUTCHER, M. D., CAMINITI, R.,
AND MASSEY, J. T. The representation of movement direction in the motor
cortex: single cell and population studies. In: Dynamic Aspects of Neocortical Function, edited by G. M. Edelman, W. E. Gall, and W. M. Cowan.
New York: Wiley, 1984, p. 501–524.
GEORGOPOULOS, A. P., KETTNER, R. E., AND SCHWARTZ, A. B. Primate mo-
01-26-99 09:03:21
neupal
LP-Neurophys
Downloaded from http://jn.physiology.org/ by 10.220.32.246 on July 31, 2017
track periods may well reflect a neural substrate for such
nonstandard mapping.
Even though the task operationally consisted of a standard
mapping of stimulus direction onto the same movement direction, the cortical processing may not make such a distinction. Common neural mechanisms may be used to solve both
standard and nonstandard mapping. Consider for example
a rifleman aiming at a moving target. The correspondence
between the gun barrel and the target might be considered by
the unskilled shooter to be a problem of standard mapping.
However, the experienced rifleman considers ‘‘windage and
elevation,’’ both nonstandard mappings designed to counter
the influence of wind and gravity. Standard mapping may
be treated as a subset of nonstandard mapping. The present
findings of differences in the PD between the cue and track
periods and the rotations of the PD trajectories during the
cue period argue against simple directional mapping and for
nonstandard mapping. The changes in the PD across the
trial provide a hypothetical substrate for the mapping of
information about directionally congruent or noncongruent
stimuli into the direction of the upcoming movement.
893
894
M.T.V. JOHNSON, J. D. COLTZ, M. C. HAGEN, AND T. J. EBNER
/ 9k2f$$de41
J278-8
gle-unit activity during sinusoidal tracing. J. Neurophysiol. 68: 528–541,
1992.
SCHWARTZ, A. B. Motor cortical activity during drawing movements: population representation during sinusoidal tracing. J. Neurophysiol. 70: 28–
36, 1993.
SCOTT, S. H. AND KALASK A, J. F. Motor cortical activity is altered by
changes in arm posture for identical hand trajectories. J. Neurophysiol.
73: 2563–2567, 1995.
SCOTT, S. H. AND KALASK A, J. F. Temporal changes in the effect of arm
orientation on directional tuning of cells in monkey primary motor (M1)
and dorsal premotor (PMd) cortex during reaching. Soc. Neurosci. Abstr.
22: 1829, 1996.
SCOTT, S. H. AND KALASK A, J. F. Reaching movements with similar hand
paths but different arm orientations. I. Activity of individual cells in
motor cortex. J. Neurophysiol. 77: 826–852, 1997.
SHEN, L. AND ALEXANDER, G. E. Neural correlates of a spatial sensory-tomotor transform in primary motor cortex. J. Neurophysiol. 77: 1171–
1194, 1997a.
SHEN, L. AND ALEXANDER, G. E. Preferential representation of instructed
target location versus limb trajectory in dorsal premotor area. J. Neurophysiol. 77: 1195–1212, 1997b.
SHEPARD, R. N. AND COOPER, L. A. Mental Images and Their Transformations. Cambridge, MA: MIT Press, 1982.
SOECHTING, J. F. AND FLANDERS, M. Moving in three-dimensional space:
frames of reference, vectors, and coordinate systems. Annu. Rev. Neurosci. 15: 167–191, 1992.
STEINMETZ, M. A., MOTTER, B. C., DUFFY, C. J., AND MOUNTCASTLE, V. B.
Functional properties of parietal visual neurons: Radial organization of
directionalities within the visual field. J. Neurosci. 7: 177–191, 1987.
TANJI, J. AND EVARTS, E. V. Anticipatory activity of motor cortex neurons
in relation to direction of an intended movement. J. Neurophysiol. 39:
1062–1068, 1976.
TURNER, R. S., OWENS, J.W.M., AND ANDERSON, M. E. Directional variation
of spatial and temporal characteristics of limb movements made by monkeys in a two-dimensional work space. J. Neurophysiol. 74: 684–697,
1995.
WEINRICH, M. AND WISE, S. P. The premotor cortex of the monkey. J.
Neurosci. 2: 1329–1345, 1982.
WEINRICH, M., WISE, S. P., AND MAURITZ, K.-H. A neurophysiological
study of the premotor cortex of the rhesus monkey. Brain 107: 385–414,
1984.
WISE, S. P. The primate premotor cortex: past, present and preparatory.
Annu. Rev. Neurosci. 8: 1–19, 1985.
WISE, S. P., BOUSSAOUD, D., JOHNSON, P. B., AND CAMINITI, R. Premotor
and parietal cortex: corticocortical connectivity and combinatorial computations. Annu. Rev. Neurosci. 20: 25–42, 1997.
WISE, S. P., DI PELLEGRINO, G., AND BOUSSAOUD, D. Dissociation of visuomotor from sensory signals. J. Neurophysiol. 68: 969–972, 1992.
WISE, S. P., DI PELLEGRINO, G., AND BOUSSAOUD, D. The premotor cortex
and nonstandard sensorimotor mapping. Can. J. Physiol. Pharmacol. 74:
469–482, 1996.
WISE, S. P. AND MAURITZ, K.-H. Set related neuronal activity in the premotor cortex of rhesus monkey: effects of changes in motor set. Proc. R.
Soc. Lond. B Biol. Sci. 223: 331–354, 1985.
ZHANG, J., RIEHLE, A., REQUIN, J., AND KORNBLUM, S. Dynamics of single
neuron activity in monkey primary motor cortex related to sensorimotor
transformation. J. Neurosci. 17: 2227–2246, 1997.
01-26-99 09:03:21
neupal
LP-Neurophys
Downloaded from http://jn.physiology.org/ by 10.220.32.246 on July 31, 2017
tor cortex and free arm movements to visual targets in three dimensional
space. II. Coding of the direction of movement by a neuronal population.
J. Neurosci. 8: 2928–2937, 1988.
GEORGOPOULOS, A. P., LURITO, J., PETRIDES, M., AND SCHWARTZ, A. B.
Mental rotation of the neuronal population vector. Science 243: 234–
236, 1989b.
GODSCHALK, M., LEMON, R. N., NIJS, H.G.T., AND KUYPERS, H.G.J.M.
Behaviour of neurons in monkey peri-arcuate and precentral cortex before
and during visually guided arm and hand movements. Exp. Brain Res.
44: 113–116, 1981.
JOHNSON, M.T.V., COLTZ, J. D., AND EBNER, T. J. Encoding of target direction, speed, and velocity in the discharge of premotor and primary motor
cortical neurons. Soc. Neurosci. Abstr. 23: 1553, 1997.
JOHNSON, M.T.V., COLTZ, J. D., HAGEN, M. C., AND EBNER, T. J. Evidence
for visuo-motor processing in the directional firing of motor cortical
neurons. Soc. Neurosci. Abstr. 22: 12, 1996a.
JOHNSON, P. B., FERRAINA, S., BIANCHI, L., AND CAMINITI, R. Cortical
networks for visual reaching: physiological and anatomical organization
of frontal and parietal lobe arm regions. Cereb. Cortex 6: 102–119,
1996b.
KALASK A, J. F., CAMINITI, R., AND GEORGOPOULOS, A. P. Cortical mechanisms related to the direction of two-dimensional arm movements: relations in parietal area 5 and comparison to motor cortex. Exp. Brain Res.
51: 247–260, 1983.
KALASK A, J. F., COHEN, D.A.D., HYDE, M. L., AND PRUD’HOMME, M. A
comparison of movement direction-related versus load direction-related
activity in primate motor cortex, using a two-dimensional reaching task.
J. Neurosci. 9: 2080–2102, 1989.
KALASK A, J. AND CRAMMOND, D. Cerebral cortical mechanisms of reaching
movements. Science 255: 1517–1523, 1992.
KORNBLUM, S., HASBROUCQ, T., AND OSMAN, A. Dimensional overlap:
cognitive basis for stimulus-response compatibility—a model and taxonomy. Psychol. Rev. 97: 253–270, 1990.
LURITO, J. T., GEORGAKOPOULOS, T., AND GEORGOPOULOS, A. P. Cognitive
spatial motor processes. 7. The making of movements at an angle from
a stimulus direction: studies of motor cortical activity at the single cell
and population levels. Exp. Brain Res. 87: 562–580, 1991.
MARDIA, K. V. Statistics of Directional Data. New York: Academic, 1972.
MASON, C. R., JOHNSON, M.T.V., FU, Q.-G., GOMEZ, J. E., AND EBNER,
T. J. Temporal profile of the directional of the discharge of dorsal premotor cortical cells. Neuroreport 9: 989–995, 1998.
MILNER, A. D. AND GOODALE, M. A. The Visual Brain in Action. Oxford,
UK: Oxford Univ. Press, 1995.
MOUNTCASTLE, V. B., LYNCH, J. C., GEORGOPOULOS, A., SAK ARA, H., AND
ACUNA, C. Posterior parietal cortex in the monkey: Command functions
for operations within extrapersonal space. J. Neurophysiol. 38: 871–908,
1975.
MUSHIAKE, H., INASE, M., AND TANJI, J. Neuronal activity in the primate
premotor, supplementary, and precentral motor cortex during visually
guided and internally determined sequential movements. J. Neurophysiol.
66: 705–718, 1991.
OJAK ANGAS, C. L. AND EBNER, T. J. Purkinje cell complex and simple spike
changes during a voluntary arm movement learning task in the monkey.
J. Neurophysiol. 68: 2222–2236, 1992.
SCHWARTZ, A. B. Motor cortical activity during drawing movements: sin-