| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Department of Psychology, University of Alabama, Birmingham, Alabama, 35294; 2 Department of Physiological Optics, University of Alabama, Birmingham, Alabama, 35294
Submitted 2 October 2002; accepted in final form 22 March 2003
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONDISCLOSURESACKNOWLEDGMENTSREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONDISCLOSURESACKNOWLEDGMENTSREFERENCES |
|---|
; Gamlin and Yoon
2000; Mays 1984;
Robinson 1972;
Wurtz 1996; however, for
recent evidence to the contrary, see
Missal and Keller 2002). In
practice, however, it is common for a desired refixation to require more than
one type of eye movement. One widely studied example of this is the
interaction between saccades and vergence (see e.g.,
Enright 1984;
Kenyon et al. 1980;
Yarbus 1957;
Zee et al. 1992;
Zhou and King 1998). In recent
years, a number of behavioral studies have explored the dynamics of saccades
and vergence movements when they occur together and have found that the
resulting interaction cannot be accounted for by a simple linear summation of
signals from completely separate oculomotor subsystems
(Collewijn et al. 1995;
Enright 1984;
Erkelens et al. 1989;
Maxwell and King 1992;
Ono et al. 1978).
Specifically, two major behavioral effects have been reported, the more well
known of which is the facilitation of vergence by saccades. Peak vergence
velocity is substantially higher when the vergence movement is accompanied by
a saccade (Enright 1984;
Erkelens et al. 1989;
Maxwell and King 1992;
Ono et al. 1978).
In addition to the speeding of vergence by saccades, Collewijn et al.
(1995) reported finding, in
humans, that peak saccadic (versional) velocities of horizontal saccades are
lower when the saccade is accompanied by vergence. These authors also compared
main sequences for each eye for saccades with and without vergence and found
that peak velocity for both eyes was lower for saccades with either
convergence or divergence than for conjugate saccades of comparable size (with
the exception of the abducting eye for saccades with divergence). Saccades
with convergence were slowed more than saccades with divergence. It is
possible that input from the vergence system might cause decreases in saccade
velocity by inhibiting neurons in one or more areas known to influence saccade
velocity.
One such area is the superior colliculus (SC). Although saccade velocity is
largely determined by downstream structures, there is good evidence that the
SC influences the speed of saccades. 1) Goossens and Van Opstal
(2000a,b) used a puff of air to cause eye blinks during saccades in monkey.
These blink-perturbed saccades were of longer duration and slower velocity
than control saccades. The decrease in saccade velocity under these
circumstances was well correlated with a decrease in the firing rate of
saccade-related neurons in SC. 2) Lesions of the SC are known to
result in slower saccades (Schiller et al.
1980). 3) When lidocaine
(Hikosaka and Wurtz 1986;
Lee et al. 1988) or muscimol
(Aizawa and Wurtz 1998;
Hikosaka and Wurtz 1985;
Quaia et al. 1998) is used to
reduce the level of activity of SC neurons, saccades are slowed. 4)
The peak velocity of microstimulation-induced saccades has been shown to be
related to both the frequency of stimulation pulses
(Stanford et al. 1996) and
current strength (Van Opstal et al.
1990). 5) Edelman and Goldberg
(2001) found a correlation
between peak spike density and the firing rate of individual saccade-related
burst neurons (SRBNs). On the other hand, Sparks and Mays
(1980) showed that identical
discharges (for a given SC neuron) may be associated with a wide range of
saccades. Taken together, these results suggest that the level of activity in
SC influences, but does not explicitly encode, saccade velocity (Goossens and
Van Opstal 2000b). These observations suggest the possibility that the slowing
of saccades with vergence might be associated with a decrease in the level of
activity of SC.
Role of superior colliculus in saccade—vergence interactions
The traditional view has been that the SC is involved in the generation of
saccadic eye movements but is not involved in the production of vergence
(Wurtz 1996). Microstimulation
of SC produces conjugate saccades but not vergence
(Robinson 1972). Even if both
SCs are stimulated simultaneously, the result is not vergence but a conjugate
saccade that represents a vector average of the saccades that would have
resulted from the activation of either site alone (unpublished observations).
Although two studies in monkeys (Judge and Cumming 1986;
Mays et al. 1986) found
vergence-related neurons in the pretectum and one study in the cat has
reported finding neurons in the rostral SC that modulate with pure vergence
(Jiang et al. 1996), no
unambiguous evidence exists for the existence of vergence-related neurons in
the primate SC.
In recent years, however, a number of lines of evidence have suggested that
this view of SC may at least be oversimplified. There is evidence from several
sources suggesting that distance-related visual information may be sent to SC.
At least four studies (in cat: Bacon et al.
1998a,b;
Berman et al. 1975; opossum:
Dias et al. 1991) have shown
that superficial layer cells are sensitive to binocular disparity. Superficial
to deep connections within SC have recently been demonstrated in tree shrew
(Lee et al. 1997) and rat
(Ozen et al. 2000).
The evidence for depth-related visual input from cortical visual areas to
deep layer SC neurons is stronger. The lateral intraparietal area (LIP) is
part of the so-called "where" visual pathway that encodes both the
direction and distance of visual targets. This area also contains cells that
appear to be involved in saccade planning. Gnadt and Mays
(1995) recorded from single
LIP neurons that were selective for eye movements to specific locations in
three-dimensional space. This area sends dense projections to both frontal eye
fields and SC (Cavada and Goldman-Rakic
1989a,b;
May and Andersen 1986). Gnadt
and Beyer (1998) were able to
antidromically activate depth-related LIP neurons from deeper layers of SC.
These authors suggested that saccade-related neurons in SC might be tuned in
three dimensions.
Chaturvedi and Van Gisbergen
(1997) have demonstrated that
the saccadic system is capable of depth-specific adaptation. Using the
intra-saccadic target displacement paradigm, these authors were able to induce
opposite gain adaptation for targets closer to, or farther away from, the
plane of fixation. Previous studies have provided evidence that the mechanism
underlying saccadic adaptation involves structures upstream from the
medium-lead burst neurons (MLBs) (Deubel
1987; Frens and Van Opstal
1994; Lemij and Collewijn
1992). On the basis of these studies and their own results,
Chaturvedi and Van Gisbergen
(1997) suggested that
saccade-related neurons in SC and cerebellum might be tuned in depth.
Chaturvedi and Van Gisbergen
(1999) found that stimulation
of the intermediate layers of the caudal portion of primate SC during combined
saccade+vergence movements caused an interruption of the vergence component.
After the end of the pulse train, the vergence movement resumed until the eyes
were on target. Once again, the possibility was suggested that saccade-related
neurons in SC might be tuned in depth. According to this idea,
microstimulation fails to elicit vergence movements because these cells are
not topographically organized with respect to the vergence component.
Stimulation of any given area in SC would activate roughly equal numbers of
"saccade+convergence" and "saccade+divergence" cells,
resulting in a net command for zero vergence change. Chaturvedi and Van
Gisbergen (2000) showed that
stimulation of the rostral portion of SC suppresses vergence movements with or
without accompanying saccades.
Thus the available literature suggests two questions with regard to SC and saccade-vergence interactions: is the slowing of saccades with vergence reflected in a suppression of the discharge of saccade-related SC cells and are these cells tuned in depth? The present study was designed to address these questions.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONDISCLOSURESACKNOWLEDGMENTSREFERENCES |
|---|
Five juvenile Macaque monkeys (Macaca mulatta) were used in this study, including four males and one female. Approval for this study was granted by the Institutional Animal Care and Use Committee at the University of Alabama at Birmingham.
Surgical procedures
Each monkey first underwent a 3- to 4-wk initial training period during which it learned to enter a primate chair and take juice from a drinking tube. After this initial training period was completed, animals underwent a series of four aseptic surgical procedures to prepare them for electrophysiological recording. Intramuscular injections of ketamine (Ketalar) were used to immobilize each monkey during preoperative procedures. During surgery, animals were intubated and maintained on 1–2% isoflurane. Heart rate, blood pressure, body temperature, and respiration were monitored constantly for the duration of each surgery. Analgesics were given for 1–3 days after surgery to minimize postoperative discomfort.
The first surgery consisted of attaching a metal post to the skull to allow
the animal's head to be fixed during training and recording sessions. First,
two stainless steel strips were adjusted to match the curvature of the
animal's skull and then affixed with bone screws. Once these strips were in
place, dental acrylic was used to attach the metal post to the upturned ends
of these strips. In the second surgery, following procedures similar to those
of Judge et al. (1980); an eye
coil was implanted underneath the conjunctiva of the right eye to allow for
the precise measurement of eye movements using the magnetic search coil
technique (Fuchs and Robinson
1966). After this second surgery, animals were trained in simple
target step conjugate saccade trials. Once the animals had been given enough
practice to become proficient at this task, an eye coil was implanted in the
left eye. After recovery from this third surgery, each monkey was trained to
perform pure vergence eye movements as well as saccades accompanied by
vergence. Finally, in the fourth surgery, recording cylinders were positioned
bilaterally over 15-mm holes drilled into the skull. Dental acrylic was used
to hold the head post and recording cylinders in place. The recording
cylinders were each centered around a point 14.5 mm lateral to the midline and
1 or 2 mm rostral of ear bar zero, at a 20° angle with respect to the
sagittal plane.
Behavioral protocol and visual display
Visual targets consisted of red Maltese crosses, subtending 1.2° of
visual angle, displayed on computer monitors (90-Hz vertical refresh rate).
The angular subtense of the monitors was ±20° horizontally and
±15° vertically. Target disparities of 
10° were generated
through the use of a dichoptic system of mirrors (see
Fig. 1). Accommodative demand
was controlled through the use of movable, servo-driven lenses. To eliminate
the problem of mechanical delays associated with the movement of these lenses,
separate monitors were used for prestep and poststep stimuli for each eye (for
a total of 4 monitors). The monitors were aligned so that corresponding points
on different monitors were superimposed in the monkey's field of view. Two
cameras were aimed so that the animal's eyes could be simultaneously centered
in two video monitors to ensure proper positioning of the animal for the
visual display.
|
During recording sessions, animals sat in a primate chair with their heads
immobilized. An audible tone signaled that a trial was about to begin. All
trials began with the appearance of a fixation target, which the monkey was
required to fixate within 500 ms. For all trial types, the fixation target
remained on for 800–1,200 ms, during which time the animal had to keep
its eyes within an imaginary, square, 10 x 10° window centered on
the target. At the end of this time, the fixation target was extinguished at
the same time that a peripheral target appeared. The horizontal and vertical
coordinates of the peripheral target were manipulated to allow the movement
field of each cell to be mapped. For each cell recorded, the amplitude and
direction of the required saccades were selected to ensure that the range of
data collection would be sufficient to obtain a good representation of the
cell's movement field. This was necessary because preliminary data analysis
showed that movement fields may shift for saccades accompanied by vergence.
Thus to ensure that the data points used for comparisons came from the same
part of the cell's movement field in each condition, it was necessary to plot
movement fields for each condition for each cell. Due to limitations of the
visual display system, recording sites were chosen with the goal of finding
cells that preferred saccades of 20°.
To properly evaluate the possibility that SC burst neurons might be tuned in three dimensions, it was necessary to be able to plot movement fields for saccades without vergence, saccades with convergence, and saccades with divergence. For example, if only two trial types (saccades without vergence and saccades with convergence) had been used, a cell preferring saccades with divergence would be indistinguishable from a cell that is suppressed by all vergence movements. It was also necessary to include saccade-only trials at both near distance and far distance. On saccade+divergence trials, the eyes start out converged and then diverged. If this had been the only trial type used in which the eyes start out converged, the monkeys would quickly learn to begin to diverge their eyes even before the target step. The inclusion of saccade-only trials at near distance avoids these anticipatory responses. In addition, it is useful to determine if there is an effect of static vergence angle. Such an effect might cause cells to fire more or less vigorously for saccades with vergence, even if the vergence movement itself has no effect.
SACCADE-ONLY TRIALS AT FAR. Both the fixation and peripheral targets appeared at optical infinity. Thus this trial type required the animal to make a saccade without vergence. (Hereafter, when the phrase "saccades without vergence" is used, it refers to this trial type.)
SACCADE-ONLY TRIALS AT NEAR. Target disparities of 8° (for monkey 254) or 10° (for all other monkeys) were matched with the appropriate accommodative demand so that fixation and peripheral targets appeared at near distance and at the same apparent distance. This trial type required the animal to make a saccade without vergence.
SACCADE+CONVERGENCE TRIALS. Target disparity and accommodative demand were manipulated so that the monkey was required to converge the eyes while making a saccade. For monkey 254, a vergence demand of 8° was used. For all other monkeys, vergence demand was 10°.
SACCADE+DIVERGENCE TRIALS. Target disparity and accommodative demand were manipulated so that the monkey was required to diverge the eyes while making a saccade. Here also, a vergence demand of 8° was used for monkey 254 and 10° for all other monkeys.
For a cell to be accepted for data analysis, 
40 correctly performed
trials had to be present for saccades without vergence and a minimum of 40
correctly performed trials from at least one of the two saccade+vergence trial
types. Sufficient data were obtained for 40 cells for saccades without
vergence and for both saccade+vergence conditions. Cells for which sufficient
data were obtained in only one of the two saccade+vergence conditions (11
cells) could still be useful to investigate some of the issues relevant to
this study; therefore these cells were included in most data analyses
(exceptions will be described in the following text).
Data acquisition and electrophysiological recording
All data were acquired using custom software running on a PC 486 DOS-based
system. Horizontal and vertical eye-position data from both eyes were acquired
through the use of the magnetic search coil technique and sampled at 1 kHz.
Single-unit activity was recorded from 51 saccade-related neurons in SC, using
low-impedance (0.1–0.3 M
) tungsten microelectrodes (Microprobe).
Unit data were passed through a window discriminator and stored for off-line
analysis. An additional channel was used to store the 20-kHz analog signal
from the electrode. Electrodes were insulated with polyamide tubing and
mounted in 26-gauge steel tubing. To penetrate the dura, 21-gauge hypodermic
needles were used as guide tubes. Electrodes were held by an X-Y positioner
and advanced by a hydraulic microdrive. High-frequency magnetic interference
from the field coils was eliminated by filtering single-unit data >5 kHz.
Unit data were passed through a window discriminator and stored on compact
disks for off-line analysis.
The criteria used to verify that the electrode was located in the SC were as follows: 1) neurons were found that displayed a high-frequency burst of spikes associated with saccades of a particular vector; 2) saccadic eye movements could be elicited by microstimulation at 20 µA; 3) electrode tracks and/or marking lesions could be observed in SC sections during histological reconstruction.
Data analysis
Data were analyzed off-line using custom software and SigmaPlot for Windows 2000. Calibration errors and nonlinearities in the search coil system were corrected by fitting eye-position data with a second-order polynomial.
For each correctly performed trial, the first saccade to the peripheral target was manually selected for data analysis. The custom software used allowed the following parameters to be measured and exported to SigmaPlot for further analysis: the size of horizontal and vertical components of the saccade for each eye; peak cyclopean eye saccade velocity, peak vergence velocity, number of spikes, peak spike density (described in the following text), burst duration, latency from burst onset to saccade onset, and latency from target onset to saccade onset.
If SC neurons are tuned in three dimensions, this should be reflected in the discharge characteristics of the cell for trials near the movement field peak. Various curve fitting approaches were tested, including three-dimensional Gaussian functions and modified three-dimensional Gaussians. These methods gave a poor fit to the data for some cells on saccade+vergence trials. Because the movement field peak was of primary importance to the issues this study was designed to address, the data analysis focused on trials near the movement field peak. In general, independent groups t-test were used to compare data between the saccade without vergence at far and each of the other conditions. For number of spikes, these comparisons were performed on the 15% of trials in each condition with the largest number of spikes. For all other parameters, comparisons were performed on the 15% of trials with the highest peak spike densities.
The inverse distance three-dimensional smoothing function in SigmaPlot was used to generate movement field plots for the cyclopean eye for each condition for each cell. These plots were used to verify that the range of data collection had been sufficient to include the movement field peak.
Many of the analyses performed required a precise quantification of the timing of saccade onset and/or offset. Saccade onset was defined as the point at which cyclopean eye saccade velocity first exceeded 15°/s, whereas saccade offset was defined as the point at which this value first fell to 25°/s. For each trial, spikes were counted within a time window beginning 30 ms before saccade onset and ending at saccade offset. The size of this window was selected to exclude as much preburst activity as possible while still being large enough to include the entire burst for each trial. Independent groups t-test were used to perform pair-wise comparisons using data from the 15% of trials with the largest number of spikes.
Spike-density functions were calculated for each trial by dividing the
measuring period into 1-ms time bins and computing a gaussian pulse for each
bin in which a spike occurred. The Gaussian pulse for the k-th bin is
given by the following equation
 | (1) |
These Gaussians were then convolved to produce a continuous function
representing spike density
 | (2) |
). The spike-density function was used as a basis for the calculation of many of the parameters used in this study. Peak spike density was defined as the largest value reached by the spike-density function during a time window beginning 30 ms before saccade onset and ending at saccade offset. It was used as a second measure of each cell's level of activity in each condition. This was done to obtain a more complete characterization of each cell's discharge properties in each condition. For example, a change in the number of spikes might simply reflect differences in preburst activity, even if the saccade-related burst itself is unchanged. On the other hand, even if peak spike density is different for saccades with vergence, the number of spikes might be the same if compensatory changes in burst duration occur or if there are changes in the preburst activity. For these reasons, both number of spikes and peak spike density were used.
As noted in the preceding text, Collewijn et al.
(1995) showed that horizontal
saccades are slower, and of longer duration, in humans when the saccade is
accompanied by vergence. This suggests the possibility that SC neurons might
show prolonged bursts with lower spike frequencies. To test this possibility,
burst duration was calculated as follows: burst onset was defined as the last
point in time before the peak at which spike density rose to a threshold of
50% of peak spike density. Burst offset was defined as the first point in time
after the peak at which spike density fell to 50% of its maximum value. Thus
burst duration was equal to the number of milliseconds between these threshold
defined points. Burst duration for the 15% of trials with the highest peak
spike densities for saccades without vergence was then compared with burst
duration for the top 15% of trials in each of the other conditions.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONDISCLOSURESACKNOWLEDGMENTSREFERENCES |
|---|
The primary objective of this study was to investigate the role of the SC in saccade-vergence interactions. It was possible to obtain sufficient data from 45 cells (of 51 total) to characterize the neuronal response properties associated with saccades with convergence. Compared with saccades without vergence, 35/45 cells (78%) showed a statistically significant decrease, and 7/45 (16%) showed a significant increase in the average peak spike density for the 15% of trials with the highest peak spike densities (P < 0.05). For saccades with divergence, it was possible to obtain sufficient data from 44 cells, of which 38 (86%) showed a significant decrease in mean peak spike density (P < 0.05). Only one cell (2%) showed a significant increase in peak spike density for saccades with divergence. Averaging across all cells, saccades with convergence were associated with a 30% reduction in peak spike density, while saccades with divergence were associated with a 25% reduction. Figure 2 shows contour plots displaying movement fields for cell 01120.254 for saccades without (Fig. 2A) and with (Fig. 2B) convergence. The contour lines in this figure indicate peak spike density. Figure 3 shows single trials from cell 00430.713 for saccades without vergence (A) and saccades with convergence (B). Number of spikes and peak spike density were both suppressed for this cell for saccades with vergence. As reported in the following text, the movement fields of some neurons shifted for saccades with vergence. However, this cell showed no movement field shift for saccades with vergence; both trials are taken from the movement field peak. Figure 4A compares each cell's response for saccades with convergence to its response for saccades with divergence. Data are expressed as a percentage of no vergence (far distance) control. The horizontal and vertical reference lines indicate 100% (or no change) for divergence and convergence, respectively. From this figure, it is clear that the overwhelming majority of cells showed lower peak spike densities for both convergence and divergence. The only cell that showed a significant increase for saccades with divergence showed no change for saccades with convergence.
|
|
|
A similar suppression of discharge was observed using number of spikes as a measure of cell activity. For saccades with convergence, 33/45 (73%) showed a significant reduction in number of spikes compared with 9/45 (20%) that showed an increase. For saccades with divergence, 31/44 (70%) showed a reduction in number of spikes. Again, only one cell (2%) showed a significant increase (this was the same cell that showed a significant increase in peak spike density for saccades with divergence). Averaging across all cells, saccades with convergence were associated with a 31% reduction in the number of spikes, whereas saccades with divergence were associated with a 23% reduction. In general, as shown in Fig. 4B, the data closely resembled the results seen for peak spike density. The major difference between the results for number of spikes and peak spike density was that the minority of cells that showed enhanced responses for vergence tended to show greater increases in number of spikes than in peak spike density. This tendency primarily reflects the fact that the small number of cells that showed increased activity for saccades with vergence also tended to show weak or moderate increases in preburst activity on some saccade+vergence trials. However, there were three cells (described in the following text) that fired vigorously for vergence movements even in the absence of saccades.
In an attempt to investigate the reasons for this suppression of neuronal
activity, it is useful to search for evidence of a relationship between peak
spike density and various other factors, such as saccade velocity, saccade
latency, intrasaccadic vergence, peak vergence velocity, and vergence motor
error at the time of saccade onset. One potential problem was the fact that
the relationship between saccade vector and peak spike density is so strong
that weaker effects are likely to be hard to detect. To overcome this problem,
it was necessary to statistically remove (as much as possible) the effect of
saccade vector on peak spike density. The procedure used takes advantage of
the fact that there is a roughly linear relationship (r 
0.7)
between the peak discharge rate of SC neurons and the distance of the endpoint
of each saccade from the movement field peak (see
Fig. 5, A and
B). The data were then fit with a linear regression line.
For a few cells, the range of data collection extended far enough beyond the
edge of the movement field that a floor effect was observed. Therefore if peak
spike density was <150 for all trials over a range of 2° along the
abscissa (distance from peak), trials over this range were not considered for
purposes of fitting the regression line for computation of normalized peak
spike density. For each trial, normalized peak spike density was calculated by
the following equation
 | (3) |
|
An added benefit of this procedure was that it provided an independent means of measuring the suppression of neuronal activity in SC. To do this, the preceding normalization procedure was repeated for saccades with vergence with the following difference: peak spike density for each trial was divided by the value of peak spike density that would have been expected (based on distance from peak) for a trial in the no vergence condition. In other words, predicted peak spike density was taken from the regression analysis for the peak spike density–distance from peak relationship for the no vergence condition. This made it possible to express peak spike density for each trial as a percentage of what would have been expected for saccades without vergence. Calculating the mean of this value across all trials gave a measure of the amount of suppression of neuronal activity for the cell across the entire movement field. Figure 6 shows that this measure agreed well with data obtained by averaging data for the 15% of trials in each condition that showed the highest peak spike densities for each cell. On the x axis is plotted the suppression index for each cell; this was calculated by dividing the mean peak spike density for the top 15% of trials in the vergence condition (usually saccades+convergence) by the mean peak spike density for the top 15% of trials in the no vergence condition and multiplying the result by 100. On the y axis is plotted the average value (for each cell) of peak spike density, normalized with respect to saccades without vergence. For this analysis, the correlation coefficient was 0.75.
|
Effect of static vergence angle
It is possible that the suppression of neuronal activity on saccade+vergence trials might, in part, reflect a dependence on static vergence angle. To test this possibility, data were collected from 28 cells for saccades without vergence at far distance (optical infinity) and near distance (eyes converged by 8° for monkey 254 and 10° for all other monkeys). Fourteen cells (50%) showed significant reductions and 8 cells (29%) showed significant increases in peak spike density for saccades at near distance. Results for number of spikes were similar to those obtained for peak spike density: 9/28 cells (32%) fired significantly fewer spikes for saccades at near distance, and 6/28 (21%) fired significantly more spikes. The average change in peak spike density was 14%, and the average change in number of spikes was 22% for saccades at near distance. Figure 7 compares movement fields for cell 01201.254 for saccade-only trials at far (A) and near (B) distance.
|
Burst duration for saccades with and without vergence
Comparisons of burst duration for each cell showed that the suppression of discharge in SC neurons is not compensated for by an increase in burst duration. On the contrary, the tendency was for burst duration to decrease for saccades accompanied by vergence. In the saccade+convergence and saccade+divergence conditions, the discharge of some neurons was suppressed to the extent that the cell no longer exhibited saccade-related bursts. Therefore analyses of burst duration were only conducted on cells for which peak spike density >400 spikes/s for a minimum of 15% of trials in each condition. For saccades with convergence, 22 cells met this criterion. Of these, 10 (45%) showed significant reductions in burst duration. Two cells (9%) showed increased burst duration. These data are summarized in Fig. 8. For saccades with divergence, burst duration data were analyzed for 26 cells. Five of these (23%) showed significant decreases in burst duration. There were no cells that showed significant increases in burst duration for saccades with divergence. Static vergence angle had relatively little effect on burst duration. For this measure, 17 cells met the criterion for inclusion, with significantly shorter bursts seen in 3 cells (18%) and significantly longer bursts in 1 cell (6%).
|
SC activity related to saccadic velocity
As noted in the preceding text, Collewijn et al.
(1995) have shown that
horizontal saccades in humans are slowed by vergence. A similar reduction in
saccadic velocity in monkeys may be due, in whole or in part, to the reduction
in colliculus activity seen for saccades with vergence. The first objective
was to determine if saccades are slowed in monkeys by comparing main sequence
data for saccades with and without vergence. Results were highly variable from
one animal to another and, in some cases, even from day to day, but generally
saccades tended to be slower on saccade+vergence trials. To quantify this,
main sequence relationships were first plotted for saccades without vergence,
saccades with convergence, and saccades with divergence. Data for each
condition were fitted with a second-order regression equation and then
normalized with respect to saccades without vergence, using the following
equation
 | (4) |
|
|
Because saccadic velocity is so strongly related to saccadic size, it was
necessary to use our normalization procedure to remove the effect of saccade
size before it was possible to determine if normalized peak spike density was
related to saccade velocity. This was done using the following equation
 | (5) |
|
These results show that decreases in saccade velocity are often associated with decreases in peak spike density. Because saccades are slower when accompanied by vergence, this suggests the possibility that this effect might explain the suppression of neuronal activity on saccade+vergence trials. However, the reduction in peak spike density was, in general, substantially greater than would be expected on the basis of the reduction in saccade velocity. To test this observation quantitatively, the relationship between normalized peak saccade velocity and peak spike density (normalized with respect to saccades without vergence) was plotted for saccades with and without vergence. This analysis was only performed on cells that showed a significant suppression of peak spike density for saccades with vergence (n = 37). The two regression lines were found to be significantly different for 73% of cells on which this analysis was performed. Thus the suppression of SC activity was usually greater than would be predicted on the basis of the velocity effect alone. Figure 10B shows these data for one typical cell.
Relationship between vergence parameters and reduced activity
The preceding results indicate that factors other than saccade velocity
must be largely responsible for the reduction in spike activity on
saccade+vergence trials. It is possible that input from the vergence system
might inhibit SC neurons. If this is the case, then the amount of suppression
observed on any given trial should be related to vergence motor parameters.
The relationship between peak vergence velocity and normalized peak spike
density was investigated. Peak vergence velocity, however, is correlated with
peak saccade velocity (C. Busettini and L. E. Mays, personal communication).
An example of this relationship is shown in
Fig. 11. To determine if peak
vergence velocity is related to peak spike density, it was necessary to
compute normalized peak vergence velocity, using the following equation
 | (6) |
|
The possibility of a relationship between intrasaccadic vergence (the vergence change during the saccade) and peak spike density was also investigated. Because the disparity of the visual target was held constant between trials, a dependence of spike activity on intrasaccadic vergence would provide evidence that the SC is, either directly or indirectly, inhibited by input from the vergence system. A significant negative correlation was found for 12/50 cells (24%). Two cells (4%) showed a significant positive correlation. Across all cells, the average value was r = –0.12. Thus the relationship between intrasaccadic vergence and firing rate appears to be very weak. For saccades with convergence, an average of 34% of the vergence change occurred during the saccade; for saccades with divergence, intrasaccadic vergence accounted for an average of 39% of the total vergence change.
Another possibility is that the activity of SC burst neurons might be affected by vergence motor error at the time of saccade onset (C. Busettini, personal communication). (Vergence motor error is equal to the difference between current and desired vergence angle.) The relationship between this variable and normalized peak spike density was examined. A significant positive correlation was found for 8/50 cells (16%), and a significant negative correlation was found for 3/50 cells (6%). Across all cells, the average value was r = 0.05. Thus the relationship between vergence motor error at saccade onset and neuronal activity was weak.
Neuronal activity related to visual factors
Although the data provide evidence that the activity of some SC neurons is
related to vergence motor parameters, no evidence was found to suggest that
this is the case for the majority of cells. This suggests the possibility that
the observed suppression of activity in SC neurons might primarily reflect the
fact that the visual target is outside the plane of fixation, instead of the
occurrence of a vergence movement, per se. If this is the case, most SC
neurons should fire less vigorously whenever the task calls for a vergence
change regardless of whether a vergence movement coincides with the saccade.
On the other hand, if the suppression of discharge is exclusively due to
inhibition from vergence-related areas, then SC neurons should show no
difference in peak spike density when there is no difference in intrasaccadic
vergence regardless of where the visual target is. Although the monkeys are
generally very good at performing these eye-movement tasks, there were some
trials for which the eye movement that occurred was inappropriate for the
trial type. For example, the animals would occasionally make a conjugate
saccade on a trial that called for a combined saccade+vergence movement. In
other trials, the animal would make both a saccade and a vergence movement but
at different times. In either case, the saccade ended up being a conjugate
saccade not the combined saccade+vergence movement that was required. This
allowed a comparison of normalized (with respect to saccades without vergence)
peak spike density for trials in which there were differences in the disparity
and accommodative demand of the visual target but no difference in the
vergence change during the saccade. To do this, data in each condition were
sorted by intrasaccadic vergence. If the range of intrasaccadic vergence for
the saccade+vergence condition overlapped with that for the no vergence
condition, then the largest region of overlap for which the difference in mean
intrasaccadic vergence was 0.5° was selected as the region for
comparison (across all 15 cells, the average difference in mean intrasaccadic
vergence for the selected regions of overlap was 0.2°). If this region of
overlap consisted of six or more trials in each condition, then
t-tests were performed on the mean normalized (with respect to
saccades without vergence) peak spike density. Care was taken to ensure that
the saccadic component of the movement was accurate even though the vergence
component was not. If the suppression of SC activity is primarily due to the
fact that the visual target is outside the plane of fixation, there should be
a fairly good match between the amount of suppression found for this missed
trials analysis and the amount of suppression found for all trials. As
Fig. 12 shows, this is the
case. The suppression index is calculated by dividing the average value of
normalized (with respect to saccades without vergence) peak spike density for
saccades with vergence by average normalized peak spike density for saccades
without vergence and multiplying the result by 100. On the x axis is
plotted the suppression index for the missed trials; on the y axis is
plotted the suppression index for all trials. The correlation was equal to
0.75 (P = 0.0012). As this figure shows, the suppression of neuronal
activity for most SC burst neurons occurs when the visual target is outside
the plane of fixation regardless of whether or not the saccade is accompanied
by vergence.
|
SC neurons that discharge for vergence in the absence of saccades
We recorded from three cells that appeared to have both saccadic and
vergence responses on the same cell. These cells burst for saccades and, as
shown in Fig. 13, had
saccade-related movement fields similar to those of other SC cells. All three
cells displayed very similar discharge properties.
Figure 14 shows single trials
from cell 00310.713. In Fig.
14A, the monkey makes a saccade without vergence. In such
trials, the cell's response was typical of saccade-related burst neurons in
SC: a crisp, high-frequency burst of spikes begins 
20 ms before saccade
onset. Figure 14B
shows a trial in which the animal was required to make a convergence movement
without a saccade. The cell began to discharge shortly before the onset of the
vergence movement. Peak spike density tended to be 200 spikes/s. The
offset of this discharge was loosely associated with the offset of the
vergence movement. As Fig.
14C shows, the cell's response was weak or absent when
the animal diverged its eyes without a saccade. When the task called for a
combined saccade+convergence movement, the result was a combination of the
responses shown in Fig. 14, A and
B—the cell shows both a saccade-related burst and
the convergence response, although the saccade-related burst is significantly
suppressed (Fig.
14D).
|
|
Movement field shifts
Movement field shifts were sometimes observed for saccades with vergence. For most cells, these shifts were small and not significant. Average horizontal movement field shift was 1.4° for saccades with convergence and 1.0° for saccades with divergence. However, as Fig. 15 shows, shifts of several degrees were observed for some cells. Although the shifts were highly variable, some possible trends could be seen. When the horizontal shift was significant, the direction of shift was toward a larger horizontal component for 10/16 cells for saccades with convergence and 7/9 cells for saccades with divergence. Significant movement field shifts in the vertical dimension were found for 6 cells for saccades with convergence and 11 cells for saccades with divergence. These shifts were in the upward direction for 5/6 cells for saccades with convergence and 10/11 cells for saccades with divergence. No relationship was found between the suppression index and the size of the movement field shift (r = 0.04). The average eccentricity of movement field peaks for the no vergence condition was 8.3 ± 4.3 ° (mean ± SD) and ranged from 1.9 to 19.8°.
|
In addition, an attempt was made to determine if each neuron was monocular or binocular by analyzing movement field shifts separately for the two eyes in each condition. The goal was to determine if the movement fields for most cells were consistently associated with one eye or the average of the two (cyclopean eye). Unfortunately, the variability in movement field shifts made it impossible to reliably classify individual neurons as monocular or binocular.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONDISCLOSURESACKNOWLEDGMENTSREFERENCES |
|---|
One of the primary goals of the current study was to investigate the possibility that the primate SC might encode, by number of spikes and/or peak spike density, the depth component of refixations in three-dimensional space. This idea implies that individual cells should fire their most vigorous bursts only for saccades of a particular vector, accompanied by particular vergence changes. If this was the case, then the current study should have found a mixture of cells preferring saccades without vergence, saccades with convergence, and saccades with divergence. Although several cells were found that burst more strongly when the saccade is accompanied by convergence, there were no cells that displayed the pattern of discharge that would be expected of a saccade+divergence cell. One cell, 00106.941, fired more strongly for saccades with divergence, but this cell also showed an enhanced discharge for saccades with convergence. The lack of any saccade+ divergence cells in the present data argues strongly that saccade-related neurons in primate SC are not, as a rule, tuned in depth.
This would mean that another explanation would have to be found for the
finding of Chaturvedi and Van Gisbergen
(1999) that stimulation of deep
layers of SC during combined saccade+vergence movements stops the vergence
component for the duration of the stimulation. One possibility is suggested by
the three saccade-related neurons that were found that also discharged during
vergence even in the absence of a saccade. Although all three such cells
described in the present study discharged during convergence and not
divergence, C. Busettini (personal communication) has encountered cells that
discharge during divergence (and not convergence) in the absence of a saccade.
Thus although the typical saccade-related neuron in SC does not appear to be
tuned in depth, it may be that microstimulation of deep layers may specify a
command for "no vergence change" by activating equal numbers of
convergence and divergence cells.
We also found that most SC burst neurons show significant changes in number of spikes and peak spike density associated with differences in static vergence angle, although this effect was somewhat inconsistent and tended to be smaller than the suppression of neuronal activity for saccades with vergence. The fact that some cells burst more vigorously and others less vigorously when the eyes are converged indicates that the suppression of SC activity for saccades with vergence is not simply a consequence of an effect of static vergence angle. Because data were only collected for two values of target distance for the initial fixation target (far distance and near distance), it was not possible to perform a detailed evaluation of the sharpness of tuning of saccade-related SC neurons with respect to static vergence angle. However, the fact that differences of 8 or 10° vergence angle usually were associated with small differences in spike activity implies that burst neurons are very broadly tuned with respect to static vergence angle. Nonetheless, it appears that some neurons burst more vigorously when the eyes are initially converged, and others burst more vigorously when the initial fixation point is near optical infinity, which suggests the possibility that the SC may carry signals related to static vergence angle.
SC activity and the slowing of saccades with vergence
SC burst neurons exhibit, on the average, approximately a 30% reduction in
both number of spikes and peak spike density for saccades with convergence and
approximately a 25% reduction for saccades with divergence. This effect was
not compensated for by an increase in burst duration. This reduction in
activity was weakly correlated with reductions in saccade velocity. Data from
studies involving pharmacological inactivation of SC suggest that decreased
activity of the burst neurons tends to cause a slowing of saccades
(Aizawa and Wurtz 1998;
Hikosaka and Wurtz
1985,1986;
Lee et al. 1988;
Quaia et al. 1998). In
addition, decreasing the frequency of stimulation pulses results in a decrease
in the velocities of stimulation-induced saccades
(Stanford et al. 1996). Thus
it seems very likely that the slowing of saccades with vergence is partially
attributable to decreased output from SC. On the other hand, there are several
reasons to suspect that this effect is largely due to events in other
structures. First, it should be pointed out that this suppression of burst
neuron activity occurred on many days when no slowing of saccades was
observed. Second, correlations between normalized peak spike density and
normalized saccade velocity, while sometimes highly significant, were not
particularly strong. Finally, it should be kept in mind that lidocaine or
muscimol injection might be expected to suppress SC far more than the
25–30% reduction in activity described in the present study. The present
results, then, are consistent with the suggestion of Goossens and Van Opstal
(2000) that, while changes in
the activity of SC can influence saccade velocity, the actual metrics of the
saccade are determined by other structures. Thus future studies should focus
on other structures, such as the pontine paramedian reticular formation (PPRF)
and the rostral interstitial nucleus of the medial longitudinal fasciculus
(riMLF), for a possible neurophysiological explanation for the slowing of
saccades with vergence.
Causes of the suppression of SC activity
There are several possible reasons why saccade-related bursts might be less
vigorous for combined saccade+vergence movements. Many neurons in SC have both
visual and motor responses in the same cell
(Mays and Sparks 1980). One
possibility is that the visual responses might be less vigorous if the visual
target is not in the plane of fixation. This would mean that burst neuron
activity might be suppressed with vergence because of a partial or complete
loss of the visual component of the response. However, this is unlikely to be
a major cause of the observed reduction in activity for two reasons. First,
the latency of the visual response of visuomotor cells is usually different
from the latency of the motor response. Visual responses begin 70 ms
after target onset and are usually over within 120 ms. Saccade latencies in
the present data were rarely <150 ms and were often 200 ms. Because the
measuring period began 30 ms before saccade onset, it would be rare for any
visually evoked spikes to be included in the spike count. Second, robust
reductions in spike activity were often observed for cells that showed no
evidence of a visual response.
A second possibility is that the activity of SC neurons might be influenced by saccade velocity. If so, then it might be that burst activity is suppressed with vergence because one or more saccade-related areas that project to SC are calling for slower saccades. According to this idea, the suppression of SC activity is only indirectly attributable to the presence of a vergence movement. If the SC is within the local feedback loop, then it is possible that events occurring downstream might influence the level of activity in SC. While this possibility cannot be completely ruled out, the data show that this explanation is unlikely to be the primary explanation for the result reported in this study. The suppression of neuronal activity was usually too large to be solely attributable to a velocity effect. A consistent relationship was found between saccade velocity and normalized peak spike density, but the aforementioned pharmacological studies suggest that the slowing of saccades is more likely to be an effect, rather than a cause, of the suppression of SC activity.
A third possibility is that the saccade-related bursts of SC neurons may be
less intense if the visual target is outside the plane of fixation. This is
not the same thing as suggesting that the suppression is due to the loss of a
visual response. Mays and Sparks
(1980) described a class of
cells they referred to as visually triggered motor cells. As noted in the
preceding text, these cells display saccade-related motor responses but only
if a visual target is present. It should be mentioned that these cells have
been consistently reported to represent a minority of saccade-related neurons
in SC (Mays and Sparks 1980;
Mohler and Wurtz 1976),
whereas the present study found suppressed activity in 75% of cells for
saccades with vergence. Mays and Sparks
(1980); for example, found
only 3/53 burst neurons that did not burst at all unless the saccade was to a
visual target and 8/53 cells that discharged weakly for saccades in the
absence of a visual target. However, most previous studies have used strict
classification criteria for classification of a cell as a visually triggered
motor cell. Cells are usually only classified as visually triggered motor
cells if they fire weakly or not at all for saccades in the absence of a
visual target. Edelman and Goldberg
(2001) found that the
saccade-related discharge of many SC neurons is less vigorous when no visual
target is present. The level of activity for the majority of these cells was
found to be related to the latency from the offset of a visual target to the
onset of the saccade. These authors suggested that the visually triggered
motor cells described by Mays and Sparks
(1980) might be at one extreme
of a continuum of visual dependence. This suggests the possibility that the
bursts of SRBNs and build-up cells might also be influenced by characteristics
of the visual target, such as blur, disparity, size, and motion. If so, then
the suppression of neuronal activity on saccade+vergence trials might be more
attributable to a dependence of saccade-related activity in SC on
accommodative blur and/or disparity than to the presence of a vergence
movement per se.
Neuronal activity was still suppressed on trials calling for a combined saccade+vergence movement even if there was no difference in the intrasaccadic vergence between these trials and trials calling for saccades without vergence. This fact, combined with the lack of a consistent relationship between vergence motor parameters (including intrasaccadic vergence, which is a relatively direct measure of saccade-vergence interaction) and peak spike density, suggests that the activity of saccade-related neurons is suppressed on saccade+vergence trials primarily because these neurons are not activated as strongly if the visual target is outside the plane of fixation.
A fourth possibility is that the motor regions of SC might receive inhibitory input from vergence-related areas. If this is the case, then peak spike density might be correlated with one or more vergence motor parameters such as intrasaccadic vergence, peak vergence velocity, or vergence motor error at the time of saccade onset. Such effects, however, are very difficult to detect because peak spike density is very strongly related to the distance of the endpoint of the saccade from the movement field peak. The normalization procedure for peak spike density was designed to remove this effect. The data suggest that some SC neurons receive input related to vergence motor activity. Neurons that discharge in association with vergence movements in the absence of saccades appear to be present in SC but were very rarely encountered in the present study. Analysis of data from the 50 cells for which normalized peak spike density could be calculated showed that the activity of some burst neurons is correlated with one or more vergence motor parameters. These </
, the location of movement field peaks for each
cell for saccades without vergence. 
, the direction and size of movement
field shifts for saccades with vergence. In some cases, the location of
movement field peak was the same for 2 different cells. When this happens, 2
arrows may appear to come from a single data point. Data points lacking an
arrow indicate cells that showed no sign of a movement field shift. For both
the saccade+convergence (A) and saccade+divergence (B)
conditions, the saccades eliciting the most vigorous bursts tended to have a
larger horizontal component and tended to be more upward than for saccades
without vergence.