 |
INTRODUCTION |
For most of the nearly 350 years since the basal
ganglia first were distinguished by Willis (1664, 1978)
,
their role in motor control has been the subject of much speculation
but little certainty (cf. Marsden 1980). In the late
1970s and early 1980s, a possible oculomotor role for the basal ganglia
was raised by anatomic data demonstrating a dense, inhibitory,
GABAergic projection from a major output nucleus of the basal ganglia,
the substantia nigra pars reticulata (SNr), to the intermediate layers
of the superior colliculus (SC) (cf. Beckstead 1981;
DiChiara et al. 1979; Hopkins and Niessen
1976; Huerta and Harting 1984; Jayaraman
et al. 1977; Ma 1989; Vincent et al.
1978; Wurtz and Albano 1980), a major saccadic
control center (for reviews of the SC and saccadic control, see
Sparks 1986; Sparks and Mays 1990;
Wurtz and Albano 1980). To demonstrate a physiological
role for this pathway and thereby link basal ganglia studies to one of
the simplest and best-studied motor systems, Hikosaka and Wurtz
recorded from hundreds of SNr neurons in monkeys trained to make
saccades in response to visual stimuli (Hikosaka and Wurtz
1983a-d).
Hikosaka and Wurtz found a population of neurons in the lateral portion
of the SNr that tonically generated action potentials at 50-100
spikes/s but decreased this rate of activity after the presentation of
saccadic targets or before the generation of saccades in oculomotor
tasks (Hikosaka and Wurtz 1983a; for similar results in
cat, see Joseph and Boussaoud 1985). Closely related
pharmacological experiments indicated that GABAergic manipulations of
both the superior colliculus and the SNr profoundly affected the
properties of saccadic eye movements (Hikosaka and Wurtz
1985a,b; for cat, see Boussaoud and Joseph
1985). Based on these data, Hikosaka and Wurtz proposed that
the SNr may be an important component of the oculomotor system that
functions by tonically inhibiting the superior colliculus and then
releasing that inhibition before saccades (Hikosaka and Wurtz
1989).
This hypothesis has been extended by the observation that the central
region of the caudate, a principal afferent source of the SNr
(Hikosaka et al. 1993; Parent et al.
1984; Szabo 1970), also carries visual and
saccade-related signals (Hikosaka et al. 1989a-c). The
central region of the caudate is itself innervated by multiple cortical
areas, including the frontal eye fields (FEF) (Kunzle and Akert
1977), an area that carries well-studied visual and
saccade-related signals (cf. Bruce and Goldberg 1985;
Segraves and Goldberg 1987). Although the FEF is only
one possible source of caudate visual and saccade-related signals (cf.
Alexander et al. 1986; Hikosaka et al.
1989a), a broadly accepted hypothesis has emerged (cf.
Alexander et al. 1986; Hikosaka and Wurtz
1989; Kandel et al. 1991; Leigh and Zee
1991; Wurtz and Hikosaka 1986) that the SNr
specifically, and the oculomotor basal ganglia in general, lies
primarily within a FEF-SC circuit. However, a more detailed comparison
of the response properties of neurons in the FEF, SNr, and SC, which
would provide a more rigorous test of the hypothesis that the SNr
relays saccade-related signals from the FEF (via the caudate) to the
SC, is not possible until the SNr has been examined with the same
degree of quantitative detail as the FEF and SC (for detailed studies
of the FEF, see Bruce and Goldberg 1985; Bruce et
al. 1985; Segraves and Goldberg 1987; Sommer and Wurtz 1998; for the SC, see Ottes et
al. 1986; Sparks 1978; Sparks et al.
1976).
There is also a more fundamental reason to provide a rigorous
quantification of the responses of SNr neurons during saccadic tasks.
Evidence from the work of Hikosaka and Wurtz suggests that the
saccade-related decreases in the firing rates of SNr neurons are
modulated by "contextual" factors, such as whether a target location is visible or remembered or whether a saccade is made inside
or outside of a behavioral task (Hikosaka and Wurtz
1983a,c). Although the nature of this context dependence has
not been fully explored, it may provide an important clue to the role
played by the basal ganglia in movement generation (Evarts et
al. 1984). To build on the work of Hikosaka and Wurtz and to
further explore the information carried by these neurons under a
variety of conditions, a rigorous quantification of nigral response
fields is necessary.
Therefore to build on earlier descriptions of SNr response properties,
allow a comparison with other oculomotor areas, and facilitate further
explorations of the role of the SNr in saccade generation, we attempted
to provide a detailed quantitative description of the relationships
between the firing rates of SNr neurons and the horizontal and vertical
amplitude of upcoming saccades during a visually guided saccade task.
For this purpose, we recorded the activity of 72 neurons from the
primate SNr while monkeys performed a large number of trials of a
delayed saccade task. To sample a wide range of movement amplitudes and
directions, on each trial the target was chosen randomly from a wide
range of possible locations. To examine changes in neuronal activity throughout the task, on each trial firing rate was measured during five
distinct intervals.
Our quantitative analysis led to three novel physiological findings
about the SNr. First, we were able to segregate and characterize four
distinct classes of saccade-related SNr neurons. Neurons in two of
these classes had response properties similar to those described by
Hikosaka and Wurtz (1983a-d), but neurons in the other
two classes were characterized by saccade-related increases in
activity. Second, for the SNr neurons we studied, the relationship between firing rate, in each of the five measured intervals, and horizontal and vertical saccade amplitude could be well described by a
planar surface. This planar relationship was qualitatively different
from the Gaussian-like relationship between firing rate and horizontal
and vertical saccade amplitude than has been found when FEF and SC
neurons have been examined in a similar manner (cf. Bruce and
Goldberg 1985; Ottes et al. 1986). Finally, our SNr neurons were found to have high trial-to-trial variance in firing
rate. In future work, use of planar regressions to quantify the
response profile of SNr neurons should prove useful for explorations of
the effects of memory, and other behavioral contexts, on the saccade-related activity of SNr neurons.
| |
METHODS |
Four male rhesus macaques (Macaca mulatta) were used
as subjects. All animal procedures were developed in association with the University Veterinarian, approved by the New York University Institutional Care and Use Committee, and designed and conducted in
compliance with the Public Health Service's Guide for the Care and Use of Animals.
Surgical and training procedures
All surgical and training procedures were performed using
standard protocols that have been described in detail elsewhere (Handel and Glimcher 1997). Briefly, in an initial
sterile surgery performed under isoflurane inhalant anesthesia, a
prosthesis for restraining the head and a scleral search coil
(Fuchs and Robinson 1966; Judge et al.
1980) for monitoring eye position were implanted. After this,
and all other surgical procedures, animals received analgesics and
antibiotics for a minimum of 3 days. After a 6-wk delay, access to
water was controlled and subjects were trained to perform oculomotor
tasks for juice rewards.
During training and subsequent recording sessions, monkeys were seated,
with their heads immobilized, in a primate chair placed 57 in from a
tangent screen containing a grid of light emitting diodes (LEDs). These
LEDs (441) formed a grid of points, separated by 2° of visual angle,
spanning 40° horizontally and 40° vertically.
To gather the data presented in this report, each monkey was trained to
produce saccadic eye movements in response to visual stimuli in a
delayed saccade task. Each delayed saccade trial (Fig.
1) began with an audible beep. Three
hundred milliseconds later a central fixation LED, which appeared
yellow to normal human observers, was illuminated and the subject was
required to align gaze with this stimulus (±3°) within 1,000 ms. Two
hundred to 800 ms after gaze was aligned with this fixation LED, a
single yellow eccentric LED was illuminated. After a further 200- to 1,200-ms delay the fixation LED was extinguished (the GO
cue), and the subject was required to shift gaze into alignment with the eccentric target LED (±3-5°) within 350 ms. If the subject's gaze remained in alignment with the target for 350-450 ms, the trial
was considered to be performed correctly. Each correct trial was
reinforced with a 300-ms noise burst which was supplemented randomly
with fruit juice on one-third to one-fifth of trials. On each trial,
the location of the target was chosen pseudorandomly, with replacement,
from the grid of LEDs. All trials were performed under dim
illumination. Subjects performed the delayed saccade task with an
intertrial interval of 200-800 ms.
View larger version (13K):
[in this window]
[in a new window]
|
Fig. 1.
Temporal sequence of events (top left), measured intervals
(bottom left), and display appearance during a typical
delayed saccade task trial in which gaze is shifted from a central
fixation point to an eccentric target. Both the target location and the
intervals between task events varied randomly. (For details, see
METHODS).
|
|
After a monkey was trained to perform the delayed saccade task, a
second sterile surgery was performed to implant a recording chamber
allowing vertical electrode penetrations into the SNr. We
stereotaxically positioned a stainless-steel receptacle (Crist Instruments) over a 15-mm-diam craniotomy centered 4.5 mm anterior and
7.5 mm lateral to the intersection of the interaural line and the
midsaggital plane (on the left side in 3 monkeys and the right side in
1), oriented the receptacle perpendicular to the stereotaxic horizontal
plane, and fastened it to the skull with orthopedic bone screws and cement.
Microelectrode recording techniques
After the monkeys were trained to perform the delayed saccade
task and had been implanted with a recording chamber,
electrophysiological recording sessions were initiated. During each
recording session a 23-gauge guide cannula was fixed to an
x-y micropositioner and used to pierce the dura. A
paralyne-coated tungsten microelectrode (Microprobe: 0.5-2.0 M
) or
a glass-coated platinum-tipped tungsten microelectrode (Ainsworth: 10- to 15-mm exposed tip) was then advanced. Individual action potentials
were identified by time and amplitude criteria and the times of
occurrence of these action potentials were recorded.
Recording protocol
Following the method of Schultz (Wolfram Schultz, personal
communication), we located the SNr in each subject by first recording from neurons of the ventroposterior complex of the somatosensory thalamus. By systematically recording from many locations in the lateral and medial divisions of this complex and determining the location on the animal's body surface that activated each neuron, we
were able to construct a topographic map of somatosensory receptive fields of ventroposterior thalamus. Once the region containing neurons
with somatosensory receptive fields centered on the lips and mouth was
located, we searched for the SNr by vertically advancing electrodes
past these neurons and into the underlying tissue. Putative SNr
neurons, with 50-125 spikes/s tonic firing rates and oculomotor
task-related modulations in activity, typically were first encountered
~2-5 mm ventral to the deepest orofacial somatosensory responses.
These neurons were typically encountered while the electrode was
advanced for 1-2 mm, after which no further cellular activity was
apparent, presumably because the electrode had entered the cerebral
peduncle. In some of our more anterior penetrations, when the electrode
was extended ventral to neurons with orofacial somatosensory responses,
we encountered neurons that were characterized by lower, more variable
tonic rates (usually <25 spikes/s) punctuated by saccade-related
bursts of activity; the background activity in the neighborhood of
these neurons was often more irregular and less vigorous than the
background activity in the neighborhood of putative SNr neurons. These
low tonic firing rates were inconsistent with Hikosaka and
Wurtz's (1983a) descriptions of SNr neurons, as well as our
putative SNr neurons, but consistent with Matsumura et al.'s
(1992) descriptions of subthalamic neurons. Indeed, marking
lesions made at the sites of these neurons were later recovered in the
subthalamic nucleus. Thus when we encountered neurons with these
physiological properties we tentatively classified them as subthalamic
neurons and did not include them in this report.
Data analysis
SINGLE-TRIAL MEASUREMENTS.
Data analysis was a three-step process. In the first step, for each
correctly executed trial, we measured the horizontal and vertical
amplitude of the saccade that aligned gaze with the target as well as
the firing rate of the neuron during 5 intervals (Fig. 1): a 200-ms
pre-trial interval ending at the onset of the beep that
initiated the trial; a 200-ms fixation interval ending at the onset of the target LED; a 200-ms visual interval
beginning 50 ms after the onset of the target; a 150-ms movement
interval beginning 50 ms before the onset of the saccade; and a
200-ms reward interval ending at the delivery of reinforcement.
DESCRIPTION OF INDIVIDUAL NEURONS.
In the second step of data analysis, for each neuron we examined the
relationship between the firing rate during a trial and the horizontal
and vertical amplitude of the movement made at the end of the trial. To
do this, we generated five response fields for each neuron:
one response field for each measured interval. Each response field
plotted the firing rate of the neuron during the interval as a function
of the horizontal and vertical amplitude of the saccade made at the end
of each trial.
To quantify the relationship between the firing rate during an interval
and the horizontal and vertical amplitude of the saccade made at the
end of each trial, we fit each response field with both a planar model
(2-dimensional least-squares regression) and a two-dimensional Gaussian
model. To compare the efficiencies with which these two models
described the data, the proportion of total variance accounted for
(VAF) by both the planar and Gaussian fits was computed as (total
variance 
residual variance)/total variance, where variance was
defined as the sum of the squared Cartesian distances between the data
and zero.
An F test was performed to determine if each planar fit was
significantly tilted. We also examined the planar fits to determine if
they consistently under- or overestimated firing rate for any particular range of movement amplitudes and directions. To do this, we
calculated the firing rate predicted by the regression and subtracted
it from the observed firing rate of the neuron. We then averaged these
residuals into 4 × 4° bins and plotted them as a
function of horizontal and vertical movement amplitude.
The Gaussian models had six free parameters: baseline rate, peak
modulation, horizontal and vertical centers, and horizontal and
vertical standard deviations. For each response field fit with the
Gaussian model, baseline rate was constrained to lie between 0 spikes/s
and the maximum firing rate observed on any trial during that interval.
The horizontal and vertical centers of the function were constrained to
be within the range of movements we sampled (i.e., between 20 and
+20°) while the horizontal and vertical standard deviations were
constrained to lie between 4 and 40°. Parameters of the model were
estimated using a Nelder-Meade simplex iterative fit that minimized the
squared Cartesian distance between the Gaussian model and the raw data
(Matlab). The optimization procedure was run for 10,000 iterations on
each of 10 sets of initial seed parameters. Because the Gaussian fits
never significantly outperformed the linear regressions, as will be
described in RESULTS, we used the parameters of the planar
fits in subsequent stages of analysis.
For some neurons, perievent time histograms also were generated to
examine the temporal relationships between modulations in neuronal
activity and significant task events. For each perievent time
histogram, we took the average firing rate, in 25-ms bins, across many
trials ending with movements of similar amplitudes and directions and
plotted those average rates as a function of time (±SD and ±SE).
Four 400-ms histograms were generated for each neuron, centered,
respectively, on the time when the monkey aligned gaze with the central
LED, the time when the target LED was illuminated, the time when the
saccade required for reinforcement began, and the time when
reinforcement was delivered.
CLASSIFICATION OF NEURONS.
By informally examining single trials and response fields for each
neuron, we found that nigral neurons exhibited one of four basic
response profiles during the delayed saccade task: a decrease in firing
rate after target onset and/or before saccade onset on trials ending
with contraversive movements, a decrease in firing rate that began
after fixation and continued until the delivery of reinforcement on
all trials, an increase in firing rate after target onset and/or before saccade onset on trials ending with contraversive movements, and an increase in firing rate after target
onset and/or before saccade onset on trials ending with contraversive
saccades and a decrease in firing rate on trials ending with
ipsiversive saccades.
These four response profiles could be distinguished by whether firing
rate increased or decreased during each interval, and we used this
property to systematize our classification scheme. First, we computed a
measure of the tonic baseline firing rate of each neuron by calculating
the mean and standard deviation of the firing rate during the pre-trial
interval across all trials. Then for each of the other four measured
intervals, we calculated the percentage of trials in which the firing
rate during the interval was 
1.5 SD below baseline. Because baseline
firing rate was normally distributed for most neurons, this threshold
would be exceeded on only slightly >6% of trials for neurons with
random rate modulations. To identify neurons that paused in a nonrandom
manner, we therefore defined a neuron as pausing during an
interval if decreases in activity of 1.5 SD occurred on >12% of
trials (see Handel and Glimcher 1997). Similarly, we
defined a neuron as bursting during an interval if there
were significant increases in activity (1.5 SD above baseline) on
>12% of trials. Neurons then were assigned to classes based on the
intervals in which they paused and burst. Neurons that paused during
the visual or movement interval, but did not pause during all
intervals, were classified as discrete pausers. Neurons that
paused during all intervals were classified as
universal pausers. Neurons that burst during the visual or movement intervals were classified as bursters. Finally, the
small number of neurons that both paused and burst during the visual and/or movement intervals were classified as pause-bursters.
When we used alternate methods (for details, see
DISCUSSION) to sort SNr neurons into categories, we
consistently found four separable groups, and nearly all of the same
neurons were grouped together by each of the classification methods we
employed. Perhaps most importantly for this characterization of the
reticulata population, the modal characteristics of the group of
neurons assigned to each class were robust; they essentially were
unaffected by changes in the classification criteria.
POPULATION ANALYSES.
The third step of data analysis was a population level description of
SNr neurons. We described the characteristics of the neurons assigned
to each cell class by extracting the relative z intercepts,
the slope magnitudes, and slope directions from the regression planes
fit to the response fields generated for each neuron. These parameters
provided estimates of the intervals during which the average firing
rates of the neurons were modulated up or down, the degree to which
these modulations were linearly related to horizontal and vertical
saccade amplitude, and the movement directions associated with the
largest modulations. Histograms and polar plots of these values were
then prepared and these data were tested for statistical significance.
Histology
The locations of recording sites were identified histologically
in two monkeys. During the 2 wk before the animals were killed, electrolytic marking lesions were made at the sites where the activity
of single neurons was recorded. Lesions were made by passing a 5 µA
anodal current through the tip of the recording electrode for 5-10 s.
At the end of this 2-wk period, the animals were premedicated with
ketamine and then killed with an overdose of thiopental sodium. They
were perfused intracardially with a saline solution followed by 4%
paraformaldehyde in phosphate buffered saline and finally by 30%
sucrose in phosphate buffered saline. The brains were removed from the
skulls, submerged in 30% sucrose for 3 days, blocked and cut into
40-µm frozen sections. The sections then were mounted and stained
with thionine, and the anatomic locations of the lesions were
identified, photographed, and recorded on camera lucida reconstructions
of the sections.
| |
RESULTS |
We examined 72 saccade-related neurons in this study. Each neuron
was examined while subjects correctly executed 50-500 delayed saccade
trials (136 ± 80; mean ± SD). After all the neuronal data were collected, each neuron was classified as described in
METHODS. For each cell class, we present data for a single
neuron having near modal response properties. We then present a
population level analysis of all the neurons in the class.
Discrete pausers
SINGLE NEURON DATA.
Thirty-five percent of the SNr neurons we studied
(n = 25) were classified as discrete pausers. Figure
2 plots the activity of a typical
discrete pauser during two delayed saccade trials. Horizontal and
vertical eye position is plotted as a function of time above the
instantaneous firing rate of the neuron. At the end of one trial
(left), the monkey made a downward and contraversive saccade. Note that shortly after the onset of this target, firing rate
decreased to nearly zero and continued at this lower level until the
saccade. During another trial (right), at the end of which
the monkey made an upward and ipsiversive saccade, the firing rate of
the neuron remained near baseline throughout the trial.
View larger version (16K):
[in this window]
[in a new window]
|
Fig. 2.
Eye position (top) and instantaneous firing frequency of
a typical discrete pauser (bottom) are plotted as a
function of time during 2 delayed saccade trials, 1 ending with a
contraversive and downward saccade (left) and the other
ending with an ipsiversive and upward saccade (right).
Tick marks indicate the onset times of the trial (Beep), fixation
(Fix), the target (Target), the initiation cue (GO), the
movement (Move), and reinforcement (Reward).
|
|
We recorded the activity of this neuron during 123 delayed saccade
trials. From these data we generated response fields that plot the
firing rate of the neuron, coded in color, as a function of the
horizontal and vertical amplitude of the saccade made at the end of the
trial for the pre-trial, fixation, visual, movement, and reward
intervals (Fig. 3A). Note that
during the visual and movement intervals, the firing rate of the neuron
was depressed below the mean baseline level for most trials but was
consistently lower on trials terminating with downward and
contraversive saccades than on trials terminating with upward and
ipsiversive saccades.
View larger version (35K):
[in this window]
[in a new window]
|
Fig. 3.
A: response fields plotting the firing rate of a typical
discrete pauser during the pre-trial, fixation, visual, movement, and
reward intervals as function of the horizontal and vertical amplitude
of the reinforced saccade. Data were averaged into 2 × 2° bins.
B: planar fits to the response fields in
A (top) and plots of the differences
between the observed and predicted spike rates on each trial
(bottom). Planar intercepts (shown graphically as tick
marks on z axes) from left to right: 72, 66, 43, 34, and
73 spikes/s. Slope magnitudes: 0.3 , 0.1, 1.3, 1.2, and 0.5 spikes · s 1 · deg1. Slope
directions (uphill): 184, 10, 342, 4, and 189°.
|
|
In Fig. 3B the optimally fit plane for each response field
is presented above the residuals for those fits. Note that there is no
systematic spatial structure to any of the residuals, indicating that
these computed planes do not systematically misrepresent the firing
rate of the neuron. Further, in all intervals these planar fits account
for the firing rate of the neuron as well as our 6 parameter Gaussian
model did (see Table 1). In fact, planar
fits performed as well as Gaussian fits at accounting for the firing
rates of almost all the nigral neurons we analyzed, during all measured
intervals (see Table 1). Therefore we used parameters from these planar
fits to determine the intervals during which the firing rate of the
neuron was modulated, whether any modulations in neuronal activity were
correlated with the horizontal and vertical amplitude of the movement
at the end of the trial, and the magnitude and orientation of any of
these spatially dependent neuronal modulations.
To determine whether the average activity of this discrete pauser
differed from baseline during an interval, we compared the average
firing rate of the neuron (the z intercept from the planar fit) during the interval with the average firing rate of the neuron during the pre-trial interval. On average, this discrete pauser fired
at 70 spikes/s before trials began. During the visual and movement
intervals, the average firing rate of the neuron decreased substantially but returned to the baseline level during the reward interval.
To examine whether there was a systematic relationship between the
firing rate of the neuron and the horizontal and vertical amplitude of
the saccade made at the end of the trial, we determined whether there
was a significant tilt to the regression planes fit to the response
fields for each interval. In the pre-trial and fixation intervals, the
planes were not significantly tilted. However, in the visual and
movement intervals, the planar fits were significantly tilted
(P < 0.001). Finally, in the reward interval, when the
average firing rate of the neuron had returned to baseline, the planar
fit was no longer significantly tilted (P = 0.07).
We also examined the steepness and orientations of the tilts of the
best-fit planes. For both the visual and movement intervals, the
regression planes were oriented so that they sloped downhill into the
contraversive hemifield (maximum positive gradients, or
uphill slope directions, for the visual and movement
intervals: 342 and 4°, respectively, where 0° is defined as
ipsiversive) and were relatively steep. This indicates that the firing
rate of the neuron was ~50 spikes/s lower on trials terminating with the largest amplitude contraversive saccades we sampled than on trials
terminating with the largest amplitude ipsiversive saccades we sampled.
Figure 4 plots four perievent time
histograms aligned on the times at which the subject fixated the
central LED, the target LED was illuminated, the saccade aligning gaze
with the target began, and reinforcement was delivered. Note that on
trials in which the subject made contraversive movements (plotted in
black), the neuron paused after target onset and continued at this
lower rate of firing until shortly after movement onset. In contrast, the average firing rate of the neuron was largely unchanged during trials in which the subject made ipsiversive movements (plotted in
gray). It is also noteworthy that, although the average firing rate was
~70 spikes/s around the time of fixation for both sets of trials,
there was substantial trial-to-trial variance (long error bars indicate
1 SD). This high trial-to-trial variability was not limited to periods
when the cell was firing at a tonic rate but also could be seen during
periods of rate modulations.
View larger version (21K):
[in this window]
[in a new window]
|
Fig. 4.
Perievent time histograms for a typical discrete pauser during 40 trials ending with contraversive (black: horizontal target position
ranged from 10 to 20° rightward) and 37 trials ending with
ipsiversive (gray: horizontal target position ranged from 10 to 20°
leftward) saccades. Histograms plot the mean (vertices), SD (long
bars), and SE (short bars) of firing rate in 25-ms bins aligned on
fixation, target onset, movement onset, and reinforcement (for details,
see text).
|
|
POPULATION ANALYSES.
As shown in Table 1, planar linear regressions were good descriptors of
the relationship between neuronal firing rate and horizontal and
vertical saccade amplitude for all 25 neurons we classified as discrete
pausers. For each neuron we compared the average firing rate (the
z intercept from each planar fit) during the fixation,
visual, movement, and reward intervals with the average firing rate
during the pre-trial interval to determine the degree to which the
firing rates of each discrete pauser were modulated in each interval (a
property that could not be predicted from our classification criteria).
Figure 5A plots histograms of
average firing rate as a percentage of baseline for all 25 discrete
pausers. During the fixation interval, the firing rate of the neurons
remained close to baseline. Not surprisingly, average firing rate did
drop below baseline during the visual and movement intervals, though it
is noteworthy that the decreases were of similar magnitude in both
intervals. Finally, in the reward interval, the firing rates of
discrete pausers tended to return to, or slightly exceed, baseline
levels.
View larger version (23K):
[in this window]
[in a new window]
|
Fig. 5.
Parameters from the planar fits to the response fields of 25 discrete
pausers measured during the fixation, visual, movement, and reward
intervals. A: histograms of average firing rate,
estimated by the z intercepts of the planar fits,
plotted as a percentage of baseline for each of the 4 intervals.
Means ± SD (gray arrows): 96 ± 15, 78 ± 19, 74 ± 17, and 105 ± 26%. B: polar plots of planar
slopes. Mean slope magnitudes during the 4 intervals (gray circles)
were: 0.18 ± 0.11, 0.61 ± 0.35, 0.76 ± 0.58, and
0.53 ± 0.38 (SD)
spikes · s1 · deg1. Mean slope
directions (gray arrows) were: 50, 330, 318, and 199°. Slope
directions which formed a significant uniform distribution
(P < 0.05) by the Rayleigh test are marked with an
asterisk.
|
|
To determine whether these changes in average neuronal activity varied
with horizontal and vertical saccade amplitude and/or with horizontal
and vertical target position, we first determined whether the
regression planes in each interval were significantly tilted. During
the fixation interval when no relationship between firing rate and the
horizontal and vertical amplitude of the, as yet, unspecified movement
could be expected, the tilt was significant (P < 0.05)
for only 1 of the 25 neurons (4%). However, planar regressions were
significantly tilted for 20 neurons (80%) during the visual interval,
19 neurons (76%) during the movement interval, and 14 neurons (56%)
during the reward interval.
Next, to examine the orientations and magnitudes of these spatially
tuned responses, we generated polar plots (Fig. 5B) of the
uphill directions and magnitudes of the slopes of the regression planes
from each interval for all 25 discrete pausers. In the fixation
interval, the tilts of the regression planes were, unsurprisingly, shallow and were not consistently oriented in any direction
(P > 0.8 using the Rayleigh test for the unimodal
distribution of a circular variable) (cf. Batschelet
1981). In the visual interval, however, the regression planes
had steeper tilts, so that, on average, there was an approximately
25-spikes/s difference in response rate for movements at the highest
and lowest points on each planar fit within our sampling range.
Moreover, the tilts of these planes were consistently oriented in the
same direction from neuron to neuron (P 
0.05). The
average planar fit was oriented so that the uphill direction
was ipsiversive and slightly upward (330°), indicating that, on
average, these discrete pausers were least active before
large-amplitude contraversive, and slightly downward, movements. During
the movement interval, neuronal responses also were modulated as a
function of horizontal and vertical saccade amplitude. The tilts of the
planar fits for this interval were even steeper but the orientations of
these planes were more variable. The mean uphill direction (318°) was
similar to that seen in the visual interval, but the slope directions
did not form a significant unimodal distribution (Rayleigh P
0.78). Finally, during the reward interval, the average tilt of the
planes was shallower, and the slope directions formed a unimodal
distribution (Rayleigh P 0.05) around an average uphill
direction of 199°, that is, into the opposite horizontal hemifield
from the direction observed during the visual and movement intervals.
Thus during the reward interval, discrete pausers tended to have ~20
spikes/s higher firing rates after 20° contralateral movements than
after 20° ipsilateral movements. Note that the estimates of mean
firing rate as percentages of baseline during this interval were often >100% because, after pausing, the firing rates of many discrete pausers transiently increased. For these neurons, the same
large-amplitude contralateral movements that produced the deepest
visual and movement interval pauses also were followed by the largest
reward interval increases in activity.
These population analyses indicate that the firing rates of the group
of neurons we classified as discrete pausers tended to decrease during
the interval between target onset and movement onset and that these
decreases were more substantial during trials ending with contraversive
movements than during trials ending with ipsiversive movements.
Universal pausers
SINGLE NEURON DATA.
Twenty-four percent of the SNr neurons we studied (n = 17) were classified as universal pausers. Figure
6 plots the activity of a typical
universal pauser during two delayed saccade trials; one ending with a
contraversive saccade and the other with an ipsiversive saccade. During
both trials, the neuron fired action potentials at a high rate until
the monkey aligned gaze with the central LED. After this fixation, the
firing rate of the neuron decreased and remained below baseline until
the trials ended. Note that the instantaneous firing frequency of the
neuron did not decrease smoothly after fixation but became highly
variable, although in both trials the neuron was practically inactive
just before the movement and around the time reinforcement was
delivered.
View larger version (16K):
[in this window]
[in a new window]
|
Fig. 6.
Activity of a typical universal pauser during 2 delayed saccade trials.
Plots are constructed similarly to those in Fig. 2.
|
|
Figure 7 plots the best-fit planes and
residuals for the five measured intervals. Note that there was very
little spatial structure apparent in the residuals in any interval,
indicating that these planes did not systematically misrepresent the
firing rate of the neuron. Furthermore in all intervals, these planar fits accounted for the firing rate of the neuron as well as our Gaussian model (Table 1).
View larger version (17K):
[in this window]
[in a new window]
|
Fig. 7.
Surfaces derived from planar fits, as well as associated residuals, for
the response fields of a typical universal pauser. Plots were
constructed similarly to those in Fig. 3B. Planar
intercepts (estimates of mean firing rate): 95, 77, 64, 18, and 33 spikes/s. Slope magnitudes: 0.1, 0.1, 0.0, 0.3, and 0.3 spikes · s1 · deg1. Slope
directions (uphill): 317, 5, 331, 198, and 353°.
|
|
For each interval, we determined whether the activity of the neuron was
modulated from baseline by examining the average firing rate across all
trials as determined by the linear fits. After the monkey fixated the
central LED, the average firing rate dropped by ~20 spikes/s and
continued to decrease until around the time of the saccade, remaining
substantially below baseline during the reward interval. We also used
the planar fits to determine whether there was a systematic
relationship between these modulations and the horizontal and vertical
amplitudes of the movements produced at the end of the trials. We found
that the tilt of the regression plane nearly reached significance in
the movement interval (P 0.05) and did reach
significance during the reward interval (P < 0.001).
To determine how strongly the firing rate of this universal pauser was
modulated by the horizontal and vertical amplitude of the movement
during these intervals, we examined the steepness of the regression
slopes. We found that the tilts of the regression slopes were
relatively shallow during both these intervals. Thus the modulation in
activity that we observed was essentially independent of the horizontal
and vertical amplitude of the movement produced at the end of the trial.
Because, during all intervals, the activity of this universal pauser
was not substantially modulated by the horizontal and vertical
amplitude of the movement, we averaged the firing rates from all 151 delayed saccade trials to produce the four perievent time histograms
shown in Fig. 8. Note the gradual decline
in firing rate that begins after fixation onset and becomes more
pronounced before movement onset. Both during periods when the average
firing rate of the neuron was near baseline and during periods when the average firing rate was significantly below baseline, the
trial-to-trial variance in firing rate for this universal pauser was
substantial.
View larger version (12K):
[in this window]
[in a new window]
|
Fig. 8.
Perievent time histograms for a typical universal pauser plotting the
average firing rate with SD and SE during all 151 trials, regardless of
saccadic amplitude and direction. Plots were constructed similarly to
those in Fig. 4.
|
|
POPULATION ANALYSES.
Although our definition of universal pausers ensures that the
average firing rates of these neurons decrease below baseline during
all intervals, the magnitude of these decreases cannot be predicted by
our classification criteria. The average depths of the decreases we did
observe can be seen in the histograms in Fig.
9A that plot mean firing rates
as a percentage of baseline for all 17 universal pausers. Although the
firing rates of all universal pausers decreased from baseline during
all postbaseline intervals, these decreases tended to be smallest
during the fixation interval. In the visual, movement, and reward
intervals, the decreases in activity tended to be more substantial; in
all three of these intervals, the mean firing rate of universal pausers
dropped to half of the baseline value.
View larger version (21K):
[in this window]
[in a new window]
|
Fig. 9.
Parameters from the planar fits to the response fields of 17 universal
pausers during the fixation, visual, movement, and reward intervals.
Plots were constructed similarly to those in Fig. 5. A:
histograms of planar intercepts plotted as percentage of baseline.
Means ± SD (gray arrows): 73 ± 17, 55 ± 22, 45 ± 25, and 51 ± 24%. B: polar plots of planar slopes.
Mean magnitudes (gray circles): 0.26 ± 0.14, 0.46 ± 0.30, 0.58 ± 0.42, and 0.43 ± 0.25 spikes · s1 · deg1. Mean directions
(gray arrows): 80, 324, 30, and 238°.
|
|
To determine whether the depths of these decreases in firing rate were
modulated by the horizontal and vertical amplitude of the movement made
at the end of the trial, first we determined whether the regression
planes in each interval were significantly tilted. During the fixation
interval 18% of the planes were significantly tilted. In the visual,
movement, and reward intervals, when the activity of the neurons was
deeply suppressed and saccadic targets were visible, 47, 53, and 65%
of the planes were tilted significantly.
Figure 9B plots the direction and magnitude of the slopes of
the regression planes from each interval for all 17 universal pausers.
In the fixation interval, the regression planes tended to have
relatively shallow tilts and were not consistently oriented in the same
direction from neuron to neuron (Rayleigh P 0.25). In
the visual, movement, and reward intervals, the planes tended to be
steeper but the orientations of the regression slopes did not form
significant unimodal distributions in either the visual, movement, or
reward intervals (Rayleigh P 0.25, 0.60, and 0.72). It
is important to note, however, that the firing rates of universal pausers decreased on all trials regardless of the amplitude and direction of the saccade.
Thus although the firing rates of all universal pausers decreased
substantially during all postbaseline intervals, the size of these
decreases depended significantly on the horizontal and vertical
amplitude of the saccade made at the end of the trial in only about
half the neurons. Moreover this spatial dependence was relatively weak
in comparison with the overall decreases in rate from baseline.
Finally, the weak spatial dependence that was observed in some
universal pausers did not have a consistent orientation from neuron to neuron.
These population analyses indicate that the firing rate of the group of
neurons we classified as universal pausers tended to be suppressed
significantly below baseline during all measured intervals, that for
all universal pausers, these pauses in activity were present on all
trials, regardless of the amplitude and direction of the saccade made
at the end of the trial, and that for some universal pausers, the size
of these pauses was modulated by the amplitude and direction of the
saccade made at the end of the trial, but the strength of these
modulations was relatively weak, and the orientations of their spatial
dependence were inconsistent.
Bursters
SINGLE NEURON DATA.
Thirty percent of the neurons we studied (n = 22) were
classified as bursters. Figure 10 shows
the activity of a typical burster during two delayed saccade trials. At
the end of one trial (left), the monkey made an upward
saccade. Before the saccade, neuronal firing rate increased and
remained at an elevated level until reinforcement was delivered. During
another trial (right), at the end of which the monkey made a
downward saccade, the firing rate of the neuron remained constant.
View larger version (19K):
[in this window]
[in a new window]
|
Fig. 10.
Activity of a typical burster during 2 delayed saccade trials, 1 ending
with an upward saccade (left) and the other ending with
a downward saccade (right). Plots are constructed
similarly to those in Fig. 2.
|
|
When we examined the response fields of the neuron during each interval
(Fig. 11), we found that during the
movement and reward intervals the average firing rate of the neuron
increased substantially. To determine whether there was a systematic
relationship between the firing rate of the this neuron and the
horizontal and vertical amplitudes of the movements produced at the end
of the trials, we examined the significance of any tilt in the planar
fits. During the visual, movement, and reward intervals, the firing
rate of the neuron was correlated significantly with horizontal and
vertical saccade amplitude (visual interval: P = 0.02, movement and reward intervals: P < 0.001). To evaluate
the magnitudes and orientations of the relationship between firing rate
in these intervals and the horizontal and vertical amplitudes of the
movement made at the end of the trial, we examined the steepness and
orientations of the tilts of the regression planes for each interval.
In the visual interval the tilt of the plane, though significant, was relatively shallow. However, in the movement and reward intervals, the
regression planes were steeply tilted and were oriented so that the
uphill direction was upward and contraversive. Thus the planar fits
indicate that the firing rate of this neuron was ~60 spikes/s higher
during the movement interval and 40 spikes/s higher during the reward
interval on trials ending with large-amplitude upward contraversive
movements than on trials ending with large-amplitude downward
ipsiversive movements.
View larger version (17K):
[in this window]
[in a new window]
|
Fig. 11.
Surfaces derived from planar fits, as well as associated residuals, for
the response fields of a typical burster. Plots were constructed
similarly to those in Fig. 3B. Planar intercepts
(estimates of mean firing rate): 63, 66, 66, 83, and 86 spikes/s. Slope
magnitudes: 0.1, 0.1, 0.3, 1.4, and 0.9 spikes · s1 · deg1. Slope
directions (uphill): 78, 121, 313, 247, and 219°.
|
|
Figure 12 plots perievent histograms
that average the firing rates of the neuron during 144 trials ending
with upward saccades (in black) and 121 trials ending with downward
saccades (in gray). Note the increase in firing rate at movement onset
for trials with upward saccades that is followed by a decrease in
activity after the reward on all trials. As with other SNr neurons,
there was substantial trial-to-trial variance in the firing rate of this burster that was present in nearly all bins, regardless of the
average firing rate.
View larger version (15K):
[in this window]
[in a new window]
|
Fig. 12.
Perievent time histograms for a typical burster plotting the average
firing rate with SD and SE during 144 trials ending with upward (black:
vertical target position ranged from 10 to 20° upward) and 121 trials
ending with downward (gray: vertical target position ranged from 10 to
20° downward) saccades. Plots were constructed similarly to those in
Fig. 4.
|
|
POPULATION ANALYSES.
Figure 13A plots
histograms of average firing rates, as a percentage of baseline, for
all 22 bursters. There was a slight tendency for the firing rate of
bursters to increase after fixation. During the visual and movement
intervals, the average firing rates of bursters tended to increase by
about one-third. These increases in activity usually grew larger by the
time reinforcement was delivered, although our population of bursters
varied widely in this respect. During the fixation interval, the tilt
of the best fit plane was significant for only 9% of the bursters.
However, during each of the visual, movement, and reward intervals, the planes for 64% of our bursters were significantly tilted. Figure 13B plots the directions and magnitudes of the slopes of the
regression planes from each interval. During the fixation interval, the
regression planes had relatively shallow tilts and were not
consistently oriented in any direction (Rayleigh P 0.42). However, during the visual, movement, and reward intervals, the
tilts of the regression planes were steeper. Moreover, during these
intervals the planes tended to be oriented so that the uphill slopes
pointed into the contraversive hemifield, although the slope directions
comprised a significant unimodal distribution only in the visual
interval (Rayleigh P 0.01, 0.12, 0.81). Thus the planar
fits indicated that, during the visual, movement, and reward intervals,
on average bursters tended to fire action potentials at a rate 25-35
spikes/s higher on trials ending with large-amplitude contraversive
saccades than on trials ending with large-amplitude ipsiversive
saccades. Note that although the regression planes tended to be
oriented in the opposite direction for bursters and discrete pausers,
this is only because the regression slopes point uphill,
toward the movements associated with the smallest decreases in the
firing rates of discrete pausers and the movements with the largest
increases in the firing rates of bursters. Both classes of neurons
tended to generate the largest modulations on trials terminating with large-amplitude movements into the contraversive hemifield.
View larger version (23K):
[in this window]
[in a new window]
|
Fig. 13.
Parameters from the planar fits to the response fields of 22 bursters
during the fixation, visual, movement, and reward intervals. Plots were
constructed similarly to those in Fig. 5. A: histograms
of planar intercepts (plotted as a percentage of baseline). Means ± SD (gray arrows): 111 ± 36, 130 ± 57, 136 ± 28, and 154 ± 82%. B: polar plots of planar slopes.
Mean magnitudes (gray circles): 0.29 ± 0.32, 0.73 ± 0.52, 0.82 ± 0.62, and 0.63 ± 0. 55 spikes · s1 · deg1. Mean directions
(gray arrows): 115, 171, 151, and 228°.
|
|
These population analyses indicate that the group of neurons we
classified as bursters tended to generate increases in firing rate
after the onset of the target LED and before the beginning of the
saccade; these increases in activity were sustained until the delivery
of reinforcement; and these increases in activity tended to be larger
on trials ending with contraversive saccades than on trials ending with
ipsiversive saccades.
Pause-bursters
SINGLE NEURON DATA.
Eleven percent of the SNr neurons we studied (n = 8) were classified as pause-bursters (see METHODS). Figure
14 plots the activity of a typical
pause-burster during two delayed saccade trials. At the end of one
trial (left), the monkey made a small amplitude downward
saccade. Shortly after target onset the firing rate of the neuron
increased abruptly, reached a peak of activity just before the
movement, and continued to fire at an elevated rate until the end of
the trial. During another trial (right), at the end of which
the monkey made an ipsiversive saccade, the neuron ceased firing action
potentials shortly after the target was illuminated and did not resume
firing at the baseline rate until just before the onset of the saccade.
View larger version (19K):
[in this window]
[in a new window]
|
Fig. 14.
Activity of a typical pause-burster during 2 delayed saccade
trials, 1 ending with a contraversive and downward saccade
(left) and the other ending with an ipsiversive and
upward saccade (right). Plots are constructed similarly
to those in Fig. 2.
|
|
Note that the regression planes do not completely capture the
systematic spatial structure of the firing rate of this neuron (Fig.
15). The residual plots for the visual
and movement intervals reveal that the regression planes consistently
underestimate the firing rate of the neuron on trials terminating with
small-amplitude downward movements.
View larger version (18K):
[in this window]
[in a new window]
|
Fig. 15.
Surfaces derived from planar fits, as well as associated residuals, for
the response fields of a typical pause-burster. Plots were constructed
similarly to those in Fig. 3B. Planar intercepts
(estimates of mean firing rate): 85, 78, 82, 115, and 117 spikes/s.
Slope magnitudes: 0.1, 0.1, 1.6, 2.0, and 0.7 spikes · s1 · deg1. Slope
directions (uphill): 168, 139, 119, 121, and 40°.
|
|
During the pre-trial interval, the average firing rate of this
pause-burster was ~85 spikes/s. During the fixation and visual intervals, this average firing rate was maintained, and only during the
movement and reward intervals was the average firing rate substantially
different from baseline. Even though there was no change in the mean
firing rate during the visual interval, the firing rate of this
pause-burster was modulated after target onset on some trials as can be
seen by examining the tilt of the regression planes.
During the fixation interval, the planar fit was not significantly
tilted. However, during the visual interval, even though the average
firing rate was close to baseline, the regression plane was
significantly tilted (P < 0.001), indicating that the firing rate of the neuron was modulated during this interval, but that,
on average, the increases in firing rate on some trials were balanced
by decreases in firing rate on others. During the movement and reward
intervals, when the average firing rate of the neuron rose above
baseline, the size of the increases in activity were still dependent on
the horizontal and vertical amplitude of the movement made at the end
of the trial; the planar fits for both intervals were significantly
tilted (P < 0.001). In the visual and movement
intervals, the regression planes had relatively steep tilts and were
oriented so that the uphill direction was downward and contraversive.
During the reward interval, the orientation of the planar fit was
similar, but the tilt was not as steep.
Figure 16 plots a set of perievent time
histograms that average the firing rates during 12 trials ending with
contraversive and downward movements (in black) and 19 trials ending
with ipsiversive and upward movements (in gray). Note that the neuron
shows a marked increase in activity after contraversive target onset
and a weaker decrement in response rate after ipsiversive target onset
and that these modulations persisted until after movement onset. Like other SNr neurons, this pause-burster showed a large amount of trial-to-trial variance in firing rate as indicated by the large standard deviation in firing rate observed in each bin.
View larger version (15K):
[in this window]
[in a new window]
|
Fig. 16.
Perievent time histograms for a typical pause-burster plotting the
average firing rate with SD and SE during 12 trials ending with
contraversive and downward (black: horizontal target position ranged
from 0 to 8° right and vertical target position ranged from 6 to 16° downward) and 19 trials ending with ipsiversive and upward
saccades (gray: horizontal target position ranged from 12 to 20° left
and vertical target position ranged from 6 to 16° upward). Plots were
constructed similarly to those in Fig. 4.
|
|
POPULATION ANALYSES.
Figure 17A plots histograms
of average firing rate as a percentage of baseline for all eight
pause-bursters. During the fixation interval, the average firing rate
of the neurons remained close to baseline. Interestingly, during the
visual interval, the average firing rate was still close to the average
baseline rate, even though the firing rates of most pause-bursters were
modulated during this interval in opposite directions for ipsiversive
and contraversive movements. However, during the movement and reward intervals, the average firing rate increased. During the fixation interval, the tilts of the regression planes were not significant for
any of the eight pause-bursters. However, 75% of the regression planes
were significantly tilted during the visual interval, 75% during the
movement interval, and 63% during the reward interval. During the
fixation interval, the regression planes (Fig. 17B) had
relatively shallow tilts and were not consistently oriented in any
direction (Rayleigh P 0.19). In the visual and movement intervals, the regression planes were tilted much more steeply. The
planes tended to be oriented so that the uphill direction was
contraversive and downward although, perhaps due to the small number of
pause-bursters in our sample, the slope directions did not form a
significant unimodal distribution (Rayleigh P 0.14 and
0.78 for the visual and movement intervals, respectively). Finally,
during the reward interval, although the average firing rate of
pause-bursters still was elevated, the firing rate of pause-bursters
did not vary as strongly with the horizontal and vertical amplitude of
the saccade at the end of the trial. Moreover, when there was a
relationship, the orientation of the planar fits did not appear to be
consistently biased in any particular direction (Rayleigh P
0.50).
View larger version (23K):
[in this window]
[in a new window]
|
Fig. 17.
Parameters from the planar fits to the response fields of 8 pause-bursters during the fixation, visual, movement, and reward
intervals. Plots were constructed similarly to those in Fig. 5.
A: histograms of planar intercepts (plotted as a
percentage of baseline). Means ± SD (gray arrows): 90 ± 15, 97 ± 14, 117 ± 21, and 125 ± 27%. B:
polar plots of planar slopes. Mean magnitudes (gray circles): 0.21 ± 0.12, 1.14 ± 0.86, 1.46 ± 1.11, and 0.56 ± 0.64 spikes · s1 · deg1. Mean directions
(gray arrows): 113, 110, 80, and 6°. Note that the firing rates of
pause-bursters were modulated during the visual, movement, and reward
intervals, and firing rates were highest during trials terminating with
contraversive movements.
|
|
These population analyses indicate that the firing rates of the group
of neurons we classified as pause-bursters increased after the
presentation of targets for contraversive saccades and decreased after
the presentation of targets for ipsiversive saccades. The increases
continued until the delivery of reinforcement, while the decreases
ended before the onset of the movement and were followed by increases
in activity.
Histology
Figure 18 contains camera lucida
reconstructions of six coronal sections from two monkeys showing the
locations of 22 of the neurons included in this report. Also shown are
the locations of seven additional neurons which were classified but
were not studied with enough trials to be included in this report. Of
these 29 neurons, 9 (2 discrete pausers, 4 universal pausers, 1 burster, and 2 pause-bursters) were recorded at the sites of
electrolytic lesions. Three neurons were recorded on the same
penetrations as lesions were made but at different depths. The
remaining 17 neurons were recorded on electrode penetrations made at
the same angle and from the same starting positions as penetrations on which lesions were made. The approximate depths of neurons in this
third group were estimated by using the last thalamic somatosensory neurons encountered (presumably the ventral edge of ventroposterior thalamus) as a point of comparison between the penetration on which the lesion was made and the penetration on which the neuron of
interest was recorded.
View larger version (31K):
[in this window]
[in a new window]
|
Fig. 18.
Camera lucida drawings of the estimated locations of 29 identified
neurons from 2 monkeys, presented in coronal section. Caudal to rostral
sections are arranged from left to right.
Nine neurons were located at the site of recovered electrolytic lesions
(A, 6 and 7; B, 1; C, 2;
D, 2; E, 1 and 4; F, 1 and
2). Three neurons were recorded on the same penetrations, but different
depths, as recovered lesions (A, 1 and 4;
F, 5). The remaining 17 neurons were recorded on
penetrations made at the same angle and from the same starting location
as recovered lesions (A, 2, 3, 5, and 8;
C, 1 and 3; D, 1, 3, 4, and 5;
E, 2 and 3; F, 3, 6, 7, and 8). We
classified 12 as discrete pausers (A, 1, 4, and 5;
B, 1; C, 1 and 3; E, 2;
F, 2, 4, 5, 7, and 8), 7 as universal pausers
(A, 2, 3, and 6; C, 2; D,
1 and 2; F, 1), 6 as bursters (A, 7 and
8; D, 3; E, 3; F, 3 and
6), and 4 as pause-bursters (D, 4 and 5;
E, 1 and 4). Twenty-two of these neurons were included
in the database described in this report; 7 neurons were not
(A, 1, 2, 4, and 5; F, 3, 5, and 7). All
neurons were located within the SNr (lateral to the SNc, medial to the
LGN, and ventral to the CP), except for 2 universal pausers which were
located in the zona incerta. CP, cerebral peduncle; LGN, lateral
geniculate nucleus; nIII, oculomotor nerve; OMN, oculomotor nucleus;
RN, red nucleus; SNc, substantia nigra pars compacta; STH, subthalamic
nucleus.
|
|
In all, 12 discrete pausers, 7 universal pausers, 6 bursters, and
4 pause-bursters were localized. All but two of these neurons were
localized in the lateral portion of the SNr, just medial and ventral to
the lateral geniculate nucleus (LGN) and lateral to the substantia
nigra pars compacta (SNc). The anatomy of this region is shown in more
detail in Fig. 19, which presents one
photomicrograph from each monkey along with three SNr lesions recovered
in those sections. The remaining two neurons were both identified as
universal pausers and were localized above the subthalamic nucleus in
the zona incerta. This indicates that neurons with the properties we
describe as typical of universal pausers are distributed both inside
and outside the architectural boundaries of the SNr. Whether these
neurons form two functionally distinct groups based on their precise
response properties cannot be concluded from our data. However, it is
clear that the majority of the universal pausers we localized
anatomically (5 of 7) were within the lateral SNr. It is worth noting
that, although six bursters and four pause-bursters were localized to
the SNr, no bursters or pause-bursters were localized to the
subthalamic nucleus, suggesting that our physiological criteria for
discriminating subthalamic neurons from SNr neurons were effective.
View larger version (110K):
[in this window]
[in a new window]
|
Fig. 19.
Photomicrographs of the ventral portions of Fig. 18, B
and E. Arrows show the locations of 3 recovered lesions
in the SNr.
|
|
| |
DISCUSSION |
Description of the population of SNr neurons
CELL CLASSES.
We recorded the activity of single SNr neurons while monkeys performed
a delayed saccade task. The firing rate on each trial was measured
during five intervals. For each of these five intervals, for each
neuron, we constructed a response field which plotted firing rate as a
function of horizontal and vertical saccade amplitude. We found that,
for nearly all neurons, these response fields were planar.
From these data, we were also able to observe that the firing rates of
some SNr neurons decreased during the delayed saccade task but the
firing rates of other SNr neurons increased. By statistically determining the in
TOP
ABSTRACT
INTRODUCTION
