The Journal of Neurophysiology Vol. 79 No. 4 April 1998, pp. 1918-1930
Copyright ©1998 by the American Physiological Society

Postsaccadic Enhancement of Initiation of Smooth Pursuit Eye Movements in Monkeys

Stephen G. Lisberger

Department of Physiology, Howard Hughes Medical Institute, and W. M. Keck Foundation, Center for Integrative Neuroscience, University of California, San Francisco, California 94143

	ABSTRACT

Abstract Introduction Methods Results Discussion References

Lisberger, Stephen G. Postsaccadic enhancement of initiation of smooth pursuit eye movements in monkeys. J. Neurophysiol. 79: 1918-1930, 1998. Step-ramp target motion evokes a characteristic sequence of presaccadic smooth eye movement in the direction of the target ramp, catch-up targets to bring eye position close to the position of the moving target, and postsaccadic eye velocities that nearly match target velocity. I have analyzed this sequence of eye movements in monkeys to reveal a strong postsaccadic enhancement of pursuit eye velocity and to document the conditions that lead to that enhancement. Smooth eye velocity was measured in the last 10 ms before and the first 10 ms after the first saccade evoked by step-ramp target motion. Plots of eye velocity as a function of time after the onset of the target ramp revealed that eye velocity at a given time was much higher if measured after versus before the saccade. Postsaccadic enhancement of pursuit was recorded consistently when the target stepped 3° eccentric on the horizontal axis and moved upward, downward, or away from the position of fixation. To determine whether postsaccadic enhancement of pursuit was invoked by smear of the visual scene during a saccade, I recorded the effect of simulated saccades on the presaccadic eye velocity for step-ramp target motion. The 3° simulated saccade, which consisted of motion of a textured background at 150°/s for 20 ms, failed to cause any enhancement of presaccadic eye velocity. By using a strategically selected set of oblique target steps with horizontal ramp target motion, I found clear enhancement for saccades in all directions, even those that were orthogonal to target motion. When the size of the target step was varied by up to 15° along the horizontal meridian, postsaccadic eye velocity did not depend strongly either on the initial target position or on whether the target moved toward or away from the position of fixation. In contrast, earlier studies and data in this paper show that presaccadic eye velocity is much stronger when the target is close to the center of the visual field and when the target moves toward versus away from the position of fixation. I suggest that postsaccadic enhancement of pursuit reflects activation, by saccades, of a switch that regulates the strength of transmission through the visual-motor pathways for pursuit. Targets can cause strong visual motion signals but still evoke low presaccadic eye velocities if they are ineffective at activating the pursuit system.

	INTRODUCTION

Abstract Introduction Methods Results Discussion References

Smooth pursuit eye movements result when a moving object is selected as a target and the visual signals created by its motion are transformed into commands for moving the eyes smoothly. Much previous research on the neural mechanisms of pursuit has reduced the system to a visual-motor reflex by presenting experimental animals with a single moving target and rewarding them for tracking the target's motion. However, there is now abundant evidence that pursuit is subject to modulation by many attentional and cognitive factors (e.g., Kowler 1990). For example, monkey subjects can select between two moving targets based on a color cue (Ferrera and Lisberger 1995), whereas human subjects attempt to anticipate the next target motion in a sequence, even under conditions where the next target motion is unpredictable (Kowler and Steinman 1981).

Studies of monkeys with abnormal pursuit have provided further evidence that extraretinal influences can cause powerful modulation of the smooth eye movements evoked by a given visual stimulus. Grasse and Lisberger (1992) studied the eye movements of one monkey that appeared to have normal processing of upward image motion but difficulty initiating pursuit for upward target motion. The normal processing of upward image motion was revealed by saccades that compensated for upward target motion (e.g., Newsome et al. 1985) and by normal upward eye acceleration for upward image motion that was presented during downward target motion. Kiorpes et al. (1996) studied the pursuit eye movements of two monkeys, which, because of early-onset strabismus, had nasalward-temporalward asymmetries in the initiation of pursuit. With monocular viewing, these monkeys could not initiate pursuit strongly for temporalward target motion. However, the one monkey that was tested had symmetrical responses when nasalward and temporalward image motions were presented as perturbations of target motion during nasalward pursuit. Although both of these examples demonstrate the important effects of behavioral context and prior instructions on the movements evoked by a given sensory stimulus, neither provides a way to quantify or control those cognitive variables.

One way to approach the diversity of the conditions that can modulate the pursuit response to a given image motion is to obtain control over the modulatory processes by devising paradigms in which the response to a given visual stimulus is related consistently to tightly controlled behavioral contingencies. In one example of this approach, Schwartz and Lisberger (1994) showed that brief perturbations of target motion elicit a much larger response if presented during ongoing pursuit than if presented during fixation. The size of the response to a given perturbation is graded as a function of ongoing smooth eye velocity. These data suggested that the pursuit system can be considered as two separate processes, one for transforming image motion on the retina into commands for smooth eye acceleration and one for controlling the gain of neural transmission through these visual-motor pathways. I think of the latter process as a "switch" or "volume control" that determines whether the visual-motor pathways are on or off and how strongly they are on.

In the present study, I demonstrate another variable that consistently affects the eye velocity evoked by a given visual stimulus. I show that pursuit eye velocity is enhanced strongly immediately after a saccade that points the fovea at the target of pursuit. In contrast to the postsaccadic enhancement seen in other kinds of smooth eye movements, such as ocular following (Kawano and Miles 1986) and disparity vergence (Busettini et al. 1996), the enhancement in pursuit is caused by the saccade itself and not by the rapid slewing of the retinal image during the saccade. I propose that postsaccadic enhancement of pursuit is an example of the extraretinal control of the gain of the pursuit system and that it is one of several mechanisms for ensuring that pursuit is fully activated whenever the images from a chosen target are about to be on the fovea.

	METHODS

Abstract Introduction Methods Results Discussion References

Eye movements were recorded in three rhesus monkeys that had been trained to fixate and track small targets presented on screens in front of them. Using methods that have been described elsewhere, monkeys were implanted with hardware for head restraint and an eye coil for monitoring eye movement (e.g., Judge et al. 1980; Lisberger and Westbrook 1985). For daily experiments, monkeys were placed in specially designed primate chairs and transported to the laboratory where they received water or juice reinforcement for tracking the targets accurately. Experiments lasted 2-3 h, after which monkeys were returned to their home cages. All methods had received prior approval from the Institutional Animal Care and Use Committee at the University of California, San Francisco.

Visual stimuli were presented either on an oscilloscope or on a tangent screen. The oscilloscope (Hewlett Packard, model 1304A) was driven by the outputs from a digital signal processing board in a Pentium PC computer. Oscilloscope targets consisted of 0.4° squares that were repainted on the screen at a temporal interval of 4 ms. The oscilloscope was placed 40 cm from the monkey's eyes and subtended 32 × 26° of visual angle. The tangent screen was 114 cm from the monkey's eyes and subtended 46 × 37° of visual angle. Its targets and backgrounds were created on an optical bench and reflected off x-y pairs of mirror galvanometers (General Scanning, CCX-660) onto the back of the screen. The inputs for the mirror galvanometers were provided by the digital-to-analog outputs from a Pentium PC computer. Visual stimuli consisted of a 0.1° red circular fixation spot, a 0.5° white circular tracking target, and a 28 × 20° random dot pattern that served as a moveable background. For both visual stimulation configurations, the experimental room was dimly illuminated with incandescent lamps. The fixation spot, tracking target, and random dot pattern had luminances of 0.2, 6.6, and 3.5 cd/m².

Experiments were presented as a series of individual trials, each of which had a duration of ~2.5-3 s. Each trial began with the monkey fixating a stationary point at straight-ahead gaze. After a random-duration interval of 1,220-1,740 ms, the fixation point disappeared and a second tracking target appeared at an eccentric location. The second target began to move upward, rightward, downward, or leftward at 20°/s either immediately or after delays of 50, 100, or 150 ms. In some experiments, the background texture appeared at the same time as the fixation point and either remained stationary or underwent a simulated 3° saccade (150°/s for 20 ms) that ended 0, 50, 100, or 150 ms after the tracking target underwent step-ramp motion. Each trial ended by having the target step 1° in the direction of motion and stop for 600 ms. The monkey was rewarded with a drop of juice or water at the end of the trial if he had kept eye position within 2-3° of target position throughout the trial. Fixation requirements were suspended for 400 ms after the appearance of the tracking target, so that the monkey would not be punished for the inherent reaction times in his saccadic and smooth pursuit eye movements. The fixation interval at the end of each trial motivated the monkeys to track throughout target motion even when it was difficult for them to do so. Trials were presented in a random order. If a trial was not completed successfully the first time, it was placed at the end of the list and repeated after the other trials had been attempted. When the monkey had completed each trial successfully, the list was reshuffled and presented again. This strategy ensured that I obtained the same number of repetitions of each target motion, while keeping the monkey unaware of which target motion would occur next. In general, the monkeys waltzed through each experiment, completing >95% of trials successfully the first time.

Data were acquired during the experiment and saved on computer disk for subsequent analysis. I digitized voltages proportional to horizontal and vertical eye position and velocity at sampling rates of 1 kHz per channel. In almost all experiments and for all the data in Figs. 2-11, eye velocity was provided by an analog circuit that differentiated eye position with a band-pass of DC to 25 Hz. Horizontal and vertical target position were reconstructed after the experiment from the same set of instructions that had been used to create commands for target motion during the experiment. For data analysis, I viewed each trial on the screen of a UNIX work station and moved a mouse-controlled cursor along the eye position traces to point out the start of the first saccade and determine if it occurred between 100 and 300 ms after the appearance of the tracking target. Except when the target had moved toward the position of fixation, trials were excluded from further analysis if the first saccade occurred either before or after this interval. Note that the delay between the appearance of the tracking target and the onset of the target ramp ranged from 0 to 150 ms, so that this criterion allows for saccades that start as early as 50 ms before or as late as 300 ms after the onset of ramp target motion. Because the criterion latency applies to the start of the saccade, postsaccadic measurements could be made as late as 360 ms after the onset of ramp target motion. For each trial that had been accepted for analysis, I used the cursor to point out the beginning and end of the rapid deflection of eye velocity associated with the first saccade. The computer then measured the mean eye velocity in the last 10 ms before and the first 10 ms after this deflection, as well as the times of these measurements.

View larger version (22K): [in this window] [in a new window]   FIG. 2. Target motions used to demonstrate postsaccadic enhancement of the initiation of pursuit. A and C: examples of smooth and saccadic eye movements evoked by a target that stepped 3° to the right and immediately ramped to the right at 20°/s. In each panel, the traces are superimposed horizontal eye and target position and horizontal eye velocity. Dashed eye velocity trace in C repeats the velocity trace from A. Vertical arrow in C is placed 200 ms after the onset of target motion. B: four target motions used to vary the time between onset of smooth target motion and the 1st saccade. Number next to each pair of target and eye position traces indicates the time delay between the step and ramp of target position. Eye position traces were selected to have saccades at the same latency after the target step and, therefore, at different latencies after the onset of the target ramp. Eye velocity traces were obtained from the 25-Hz differentiator, and the vertical arrows show the beginning of the 10-ms intervals used to measure postsaccadic eye velocity.

As illustrated in Fig. 1, the use of the differentiator with a band-pass of DC to 25 Hz has the disadvantage that the rapid deflections of eye velocity outlast the saccades that caused them. As a result, our measurements of postsaccadic smooth eye velocity were made in the interval that started 30 ms after the end of the actual saccade (Fig. 1, upward arrows). To determine whether this introduced a consistent bias in our data, I repeated the basic experiment on one monkey using eye velocity records from a differentiator with a band-pass of DC to 100 Hz. Figure 1A compares the eye velocity resulting from these two differentiators for a single trial in which the target stepped 3° to the right and then ramped without further delay to the right at 20°/s. The trace obtained with the 100 Hz differentiator has more high-frequency noise, but otherwise shows the same features as that obtained with the 25 Hz differentiator. Presaccadic eye velocity is low and postsaccadic eye velocity is high. As expected, the rapid deflection of eye velocity caused by the saccade outlasts the end of the saccade (vertical dashed line) in both traces, but by a longer time with the 25-Hz differentiator. In this example, the eye velocity measured from the 100 Hz differentiator immediately after the vertical dashed line would have been higher, and not lower, than that measured from the 25-Hz eye velocity records at the time indicated by the arrow. Measurements made from the two differentiators at any time after the upward arrow would have been very similar.

View larger version (22K): [in this window] [in a new window]   FIG. 1. Comparison of eye velocity traces produced by differentiators with band-passes of 25 and 100 Hz. A: traces from a single trial. Top to bottom: traces are superimposed eye and target position, eye velocity from the 25-Hz differentiator, and eye velocity from the 100-Hz differentiator. Vertical dashed line shows the time when the saccade ended on the position trace, and vertical arrow is drawn 30 ms later at the time when eye velocity would have been measured from the 25-Hz trace. B: averages of eye position and eye velocity from the 2 differentiators aligned on the end of the saccade as determined from the eye position record. Vertical dashed line, time selected as the end of the saccade. Vertical arrow is positioned 30 ms later, at the time when eye velocity would have been measured from the output of the 25-Hz differentiator. The average eye velocity from the 100-Hz differentiator is shown as a solid trace at the bottom of the panel and as the dashed trace superimposed on the output from the 25-Hz differentiator (in the middle of the panel.).

To further analyze possible artifacts incorporated by the 100-Hz differentiator, I have done one additional analysis. For the one experiment that used both the 25- and the 100-Hz differentiator, I used the eye position records to mark the end of the first saccade for all 31 trials in which the target stepped 3° to the right and ramped without delay to the right at 20°/s. Averaging eye position and the two eye velocity traces around the end of the saccade yielded the records that appear in Fig. 1B. The superimposed traces in Fig. 1B, middle, clearly show that the rapid deflections of eye velocity during the saccade had a faster rise and an earlier return to baseline for the 100-Hz differentiator (vertical dashed line) than for the 25-Hz differentiator (solid line). However, these records give no indication that a more accurate estimate of postsaccadic eye velocity could have been obtained from the 100-Hz differentiator than from measurements made 30 ms after the end of the saccade at the time shown by the vertical arrow. The exact value of postsaccadic eye velocity would have depended on the time of the measurement, but the values would have been nearly the same if measured at the same time. Given the inevitable noise on our eye movement recordings and the uncertainty about the actual end of a saccade that occurs during pursuit, I conclude that my measurements of eye velocity, although imperfect, provide a valid estimate of the smooth pursuit eye velocity in the immediate wake of a saccade.

	RESULTS

Abstract Introduction Methods Results Discussion References

I present three sets of experiments that used different combinations of target steps and ramps. First, I demonstrate the existence of postsaccadic enhancement of pursuit when the target steps 3° eccentric on the horizontal axis and ramps upward, downward, or away from the position of fixation. I also compare the enhanced postsaccadic responses with the presaccadic eye velocity evoked when the target ramps toward the position of fixation from 3° eccentric. Second, I contrive a combination of target step and ramp so that there are examples where the direction of the saccade is purely vertical and the direction of target motion is purely horizontal to show that the postsaccadic enhancement occurs even when saccadic and smooth eye movements are in orthogonal directions. Last, I analyze the postsaccadic enhancement of pursuit when the target steps horizontally and ramps either toward or away from the position of fixation from positions up to 15° eccentric in the visual field. This stimulus configuration demonstrates postsaccadic enhancement when the saccade is in either the same or the opposite direction from target motion.

Basic description of postsaccadic enhancement of smooth pursuit

To demonstrate postsaccadic enhancement of smooth pursuit eye movements, it is necessary to show that the eye velocity at a given time after the onset of target motion depends on whether the measurement was made before or after the first saccade to the target. Figure 2A shows a typical example of the eye velocity induced by a step-ramp target motion in which the step was 3° to the right and the ramp was rightward at 20°/s. There was relatively little presaccadic eye acceleration and the saccade occurred quite early, ~140 ms after the onset of target motion. However, the immediate postsaccadic smooth eye velocity was close to the target velocity of 20°/s. Figure 2C shows a second example of the eye velocity induced by the same step-ramp target motion that was used to obtain the response in Fig. 2A. In this response, the presaccadic eye acceleration was clearly visible and eye velocity reached ~5°/s before saccade onset, which occurred 205 ms after the onset of target motion. Comparison of the eye velocity traces when the saccade latencies were long (Fig. 2C, solid line) or short (Fig. 2C, dashed line) suggests that the eye velocity at a given time after the onset of target motion indeed depends on whether the measurement is made before or after the first saccade has intervened. For this pair of responses, the eye velocity 200 ms after the onset of target motion (Fig. 2C, upward arrow) was 5°/s if measured before the saccade and 20°/s if measured after the saccade.

One way to explain the large increment in eye velocity during the saccade is to assume that a rapid smooth eye acceleration was obscured by the saccade. In Fig. 2A, for example, the eye velocity increased from 4.96°/s just before the saccade to 22.88°/s at the end of the saccade, 55 ms later. This implies that there would have been a smooth eye acceleration of 326°/s² if the saccade had not occurred. During the same 55-ms interval, the same monkey showed an average eye acceleration of 123°/s² for targets that started 3° eccentric and moved toward the position of fixation, i.e., for the most effective configuration for the presaccadic initiation of pursuit. This suggests that it is incorrect to think of the large increment in eye velocity during saccades as a normal smooth eye acceleration and, instead, it seems likely a preexisting command for pursuit is revealed in the immediate wake of a saccade.

The large increment in eye velocity immediately after the saccade has been noticed by other investigators, including Robinson (1965) and Newsome et al. (1985), who used postsaccadic eye velocities as a measure of the strength of pursuit to quantify the effects of lesions of the middle temporal (MT) visual area. The latter authors did not question why eye velocity was so much larger after the saccade but pointed out that the smooth eye velocity immediately after the first saccade was a valid probe of the visual motion processing for pursuit because it was part of the open-loop response of the pursuit system to the image motion present before the saccade. In the present paper, I ask why postsaccadic eye velocities are so much higher than presaccadic eye velocities.

Instead of relying on the natural variation in saccade latency to yield chance observations like the one in Fig. 2C, I used the experimental design illustrated in Fig. 2B to systematically vary the interval between the onset of target motion and the first saccade. Each trial began with a 3° step of target position, but the 20°/s ramp either occurred immediately after the step (trace labeled 0 ms) or was delayed by 50, 100, or 150 ms (delays of 200 ms also were used but are not illustrated here). In the first version of the experiment, reported in Fig. 2-5, the target step was 3° rightward or leftward but the ramp target motion could be rightward, leftward, upward, or downward. To analyze the data, I measured the eye velocity in the last 10 ms before and the first 10 ms after the rapid deflection caused by the first saccade. I then plotted eye velocity as a function of the interval between the time of the measurement and the onset of the ramp target motion. This yielded graphs like those in Fig. 3, where each point shows measurements made at one time from one trial and the open and filled symbols plot measurements made before and after the first saccade, respectively. In these graphs, each measurement of eye velocity is plotted as a function of the latency between the onset of ramp target motion and the time the measurement was made. Recall that I analyzed trials in which the first saccades occurred 100-300 ms after the appearance of the tracking target and that the onset of ramp target motion was between 0 and 200 ms after the appearance of the target. If, for example, the saccade occurred 150 ms after the appearance of the target and the ramp of target motion began 100 ms after the appearance of the target, then I would have obtained a presaccadic measurement of eye velocity 50 ms after the onset of target motion. This datum would plot at 50 ms in Fig. 3.

View larger version (20K): [in this window] [in a new window]   FIG. 3. Examples of analysis to demonstrate postsaccadic enhancement of the initiation of pursuit. Two graphs plot measurements of eye velocity for rightward (A) and leftward (B) target motion from 1 daily experiment on 1 monkey. Each point shows 1 measurement of eye velocity from 1 trial. Each trial provided 2 points, 1 each for the measurement of presaccadic eye velocity (open symbols) and postsaccadic eye velocity (filled symbols). As indicated in the key (upper right corner in B), the different symbol shapes show data for trials in which there were different delays between the target step and ramp. All data are plotted as a function of the time between the onset of ramp target motion and the time of the measurement. The solid traces were obtained by fitting the pre- and postsaccadic data with Eq. 1. Both rightward and leftward eye velocities have been plotted as positive numbers in this and subsequent figures. Horizontal dashed lines show 0 eye velocity. Data are from monkey I. Normalized squared errors of the fits were 3.84 and 2.70 for the data in A and B.

Inspection of the two graphs in Fig. 3 shows that at any given time after the onset of target motion, the eye velocity was consistently higher if measured after (filled symbols) rather than before (open symbols) the first saccade. The different symbols show how the delay between the target step and ramp helped to fill out the time course of pre- and postsaccadic eye velocity. When there was no delay between the target step and ramp, the postsaccadic measurements (filled square) were made after the presaccadic measurements (open square), precluding direct comparison. However, when there was a 50-ms delay between the target step and ramp, the postsaccadic measurements (filled triangle) were made at about the same time relative to the onset of ramp target motion as were the presaccadic measurements for zero delay (open square). Direct comparison reveals much higher eye velocities for post- than presaccadic eye velocity measured at the same time.

To describe these data, I fitted the pre- and postsaccadic measurements with the equation

(1)

where t is time since the onset of target motion, E(t) is eye velocity at time t, and a, b, c, and d are free parameters. I fitted the data under the constraint that the same values of a, b, and c were used for both the pre- and postsaccadic fits, but each fit could have different values of d. Thus I enforced the same time course of the fit for the pre- and postsaccadic data but allowed different amplitudes. Examples of the quality of fit are shown by the smooth curves in Fig. 3, A and B.

I assessed how well the data were fitted by the curves by computing the summed squared deviation of the actual velocities from the curves and dividing by the number of points. For Fig. 3, A and B, this normalized squared error was 3.84 and 2.70. For all other curves shown in this paper, the normalized squared error is given in the figure legend. For all 102 curves fitted to the data for this paper, the value of normalized squared error ranged from 0.97 to 7.5 with a mean of 3.08 ± 1.54 (SD) and a median of 2.68. Thus the fits shown in Fig. 3, A and B, are representative of the fits I obtained for all experiments.

In subsequent graphs, I will use the value of the smooth curve 200 ms after the onset of target motion to quantify the magnitude of pre- or postsaccadic eye velocity for a given condition. I elected to measure eye velocity at this time because the resulting values could be compared directly with previous work on the initiation of pursuit eye movements. These earlier studies typically have measured eye velocity 100 ms after the onset of pursuit or eye acceleration in the first 100 ms of pursuit. Because the onset of pursuit occurs ~100 ms after the onset of ramp target motion, values obtained 100 ms after the onset of pursuit map very easily onto measurements of eye velocity made 200 ms after the onset of target motion. This said, the exact time of the measurement doesn't affect the main comparison that I will be making, which is the ratio of post- to presaccadic eye velocity at a given time. Because of the way the fitting procedure was constrained, the only difference between the pre- and postsaccadic curves is the amplitude. Thus the ratio of post- to presaccadic eye velocity will be the same at all times.

Our use of the values of the fitted curves 200 ms after the onset of target motion to summarize our results raises an important question about the fits. Inspection of Fig. 3, A and B, reveals that the times of measurements for postsaccadic eye velocity are invariably longer than those for presaccadic eye velocity and that it was not unusual for all saccades to have latencies <200 ms. To verify that it was valid to compare curves based on data over different time intervals and that the absence of presaccadic data 200 ms after the onset of target motion wasn't a problem, I re-did all the curve fits with the restriction that postsaccadic data were included in the fits only for times up to the maximum time in the presaccadic data. The resulting amplitudes of the pre- and postsaccadic curves were nearly identical to those obtained with the full data set. Comparison by regression analysis revealed a correlation coefficient of 0.996 and a slope of 1.009.

Figure 4 plots the value of the curve fitted to the presaccadic eye velocity 200 ms after the onset of target motion as a function of that for the postsaccadic eye velocity. All the points plot below the line of slope one (dashed line), indicating that postsaccadic eye velocity at this time was always substantially larger than presaccadic eye velocity. The postsaccadic enhancement of pursuit was present whether the background was a homogenous gray screen (open circle and filled triangle) or a high-contrast random dot pattern (open diamond). Postsaccadic eye velocity was distributed over about the same range whether the target ramped horizontally away from the position of fixation (open symbols) or vertically from initial positions that were 3° rightward or leftward (filled triangle). Thus at least for this constrained set of initial positions and directions of target motion, the enhanced performance of the pursuit system did not depend on the axis of target motion. However, the presaccadic eye velocity sometimes was larger when the target moved vertically, indicating that the presaccadic pursuit was stronger under this condition than when the target moved directly away from the position of fixation.

View larger version (20K): [in this window] [in a new window]   FIG. 4. Summary of postsaccadic enhancement of the initiation of pursuit in 9 experiments on 3 monkeys. Each symbol plots the results for 1 direction of target motion. The x axis plots the values obtained by evaluating the fits of Eq. 1 to the measurements of postsaccadic eye velocity 200 ms after the onset of ramp target motion. The y axis plots the results of the same procedure for presaccadic eye velocity. The dashed line has a slope of one and would obtain if the pre- and postsaccadic eye velocity showed the same relationships to the time from the onset of target motion. Different symbols show results for different stimulus configurations. Open circle, targets started 3° eccentric on the horizontal axis and moved horizontally away from the position of fixation at 20°/s across a homogeneous gray screen. Open diamond, targets started 3° eccentric on the horizontal axis and moved horizontally away from the position of fixation at 20°/s across a stationary textured background. Filled triangle, targets started 3° eccentric on the horizontal axis and moved upward or downward at 20°/s across a homogeneous gray screen.

Lisberger and Westbrook (1985) showed that the initiation of pursuit included large presaccadic eye velocities when the initial target step arranged for the moving target to cross the position of fixation after ~100 ms. Since then I have relied heavily on this configuration of the step-ramp stimulus to probe the visual motion processing for pursuit eye movements. My discovery of postsaccadic enhancement of pursuit raises the question of the relationship between the eye velocities evoked before the first saccade for target motion toward the position of fixation and those measured after the first saccade for target motion away from the position of fixation. Figure 5 shows this comparison for two directions of pursuit in each of my three monkeys. In each panel, the solid trace plots average eye velocity as a function of time after the onset of motion for targets that stepped 3° eccentric and ramped without delay toward the position of fixation. Trials were included in these averages only if the saccade was delayed beyond the end of the average, as it usually is when the target moves toward the position of fixation from 3° eccentric. The points plot postsaccadic eye velocity for targets that started 3° eccentric and moved away from the position of fixation. In four of the six cases (Fig. 5, B and D-F) there was excellent agreement between the points and the traces, and in one case (C), the disagreement was in the latency rather than the initial eye velocity of pursuit. I conclude that presaccadic pursuit is fully or almost fully enhanced when the target moves toward the position of fixation from 3° eccentric, as if a saccade already had occurred. However, the two instances of clear latency shifts between the points and the traces (Fig. 5, A and C) raise the possibility for future investigation that there is postsaccadic enhancement of the latency as well as the gain of the visual-motor pathways for pursuit.

View larger version (22K): [in this window] [in a new window]   FIG. 5. Comparison of presaccadic eye velocity for targets that moved toward the position of fixation and postsaccadic eye velocity for targets that moved away from the position of fixation. Six graphs show the results for 2 directions of horizontal target motion in each of the 3 monkeys. In each graph, the points plot postsaccadic eye velocity as a function of the time from the onset of ramp target motion away from the position of fixation. Traces show averages of presaccadic eye velocity as a function of time from the onset of target motion for targets that moved toward the position of fixation. Horizontal dashed lines show 0 eye velocity. A-C and D-F show results for rightward and leftward target motion, respectively. Left to right: data from monkeys E, I, and K.

Postsaccadic enhancement is not due to visual stimulation during saccades

I adopted the approach of Kawano and Miles (1986) to assess whether postsaccadic enhancement of pursuit was caused by the visual stimulation during saccades or by the motor act of the saccades themselves. I conducted experiments in the presence of a background that consisted of a random dot pattern, and I arranged for the background either to remain stationary or to undergo a simulated saccade that ended 0, 50, 100, or 150 ms after the onset of step-ramp target motion (Fig. 6). This selection of delay times corresponded to simulated saccades that ended ~150, 100, 50, or 0 ms before the first saccade evoked by step-ramp target motion consisting of a 3° step and 20°/s ramp in the same direction. The simulated saccade consisted of a 20-ms ramp of background position that had an amplitude of 3° and therefore a velocity of 150°/s. Figure 7A shows that the postsaccadic enhancement was very similar in the presence of a stationary background as it had been in the experiments illustrated in Fig. 3. Figure 7B shows that postsaccadic enhancement still was present when the background executed a simulated saccade. In trials that contained a simulated saccade, I wanted to determine whether background motion enhanced presaccadic pursuit. Therefore it was not necessary to insert a delay between the step and ramp of target motion, and neither the pre- nor the postsaccadic measurements shown in Fig. 7B were made over the full range of times after the onset of target motion.

View larger version (18K): [in this window] [in a new window]   FIG. 6. Four background motions used to present simulated saccades. In all cases, there was 0 delay between the target step and ramp, but the background underwent a brief excursion (3° in 20 ms) that ended 0, 50, 100, or 150 ms after the onset of the target motion.

View larger version (38K): [in this window] [in a new window]   FIG. 7. Absence of enhancement of the initiation of pursuit in the wake of a simulated saccade. A and B: graphs for 1 representative experiment. Pre- and postsaccadic eye velocity are plotted as a function of the interval between the onset of ramp target motion and the time of the measurement. Filled square and open square, post- and presaccadic measurements, respectively; solid traces show fits of Eq. 1 to the data for pre- and postsaccadic eye velocity. In A, the textured background was present but remained stationary and different trials had delays of 0, 50, 100, or 150 ms between the target step and ramp. Normalized squared error of the fit was 3.67. In B, there was always 0 delay between the target step and ramp, but the background underwent a simulated saccade that ended 0, 50, 100, or 150 ms after the onset of the target motion. Background motion was always in the opposite direction from the target motion, simulating a saccade in the same direction as target motion. Normalized squared error of the fit was 6.20. C-F: summary graphs for 2 directions of target motion in 2 monkeys. Each graph plots presaccadic eye velocity as a function of the interval between the end of the simulated saccade and the time of the measurement. Different open symbols show cases where the simulated saccade ended at different times after the step-ramp of target motion: square, 0 ms; triangle, 50 ms; inverted triangle, 100 ms; diamond, 150 ms. In all graphs, horizontal dashed lines show 0 eye velocity. A-D show data from monkey E. E and F are from monkey I.

If the simulated saccade caused enhancement of smooth eye velocity, then the enhancement should be visible in plots of presaccadic eye velocity as a function of the time between the end of the simulated saccade and the measurement of eye velocity. Enhancement would cause presaccadic eye velocity to show a peak sometime after the end of the simulated saccade and then decay back to the values of eye velocity recorded in the absence of a simulated saccade. No such peak appears in Fig. 7, C-F, which plots the results for two directions of target motion in two of our monkeys. Identical data were obtained in a third monkey. Thus I found no evidence that pursuit was enhanced at any time in the wake of the visual stimulation caused by the saccadic motion of the background without a real saccade.

Postsaccadic enhancement of pursuit is preserved when saccade is orthogonal to direction of target motion

I used the experiment diagrammed in Fig. 8B to contrive circumstances under which the direction of the saccade would be orthogonal to the direction of target motion. In different trials, the initial step of target position took the spot 6° above or below the horizontal meridian and 3, 3.75, or 4.5° right or left. After the initial target step, there was a delay of 0, 50, 100, or 150 ms before the target ramped to the right or left at 20°/s. When the target moved toward the vertical meridian from horizontal positions of 3, 3.75, or 4.5°, a pure vertical saccade would be required if saccade latency were 150, 187.5, or 225 ms. Thus this experiment is designed to take advantage of the natural variation in saccade latency to obtain examples where the saccade was vertical and the smooth target motion was horizontal, toward the vertical meridian.

View larger version (31K): [in this window] [in a new window]   FIG. 8. Analysis of the effect of the relative directions of saccade and target motion on postsaccadic enhancement of pursuit. A: summary of results when the target started 6° up and 3.75° left and moved to the right at 20°/s. Each point plots 1 measurement of presaccadic eye velocity (open symbols) or postsaccadic eye velocity (filled symbols) as a function of the interval between the onset of target motion and the time of the measurement. Different symbols indicate trials with different delays between the target step and ramp: squares, 0 ms; triangles, 50 ms; inverted triangles, 100 ms. Solid traces, fits of Eq. 1 to the data for pre-and postsaccadic eye velocity; horizontal dashed line, 0 eye velocity. Normalized squared error of the fit was 1.56. B: diagram of target motions used to control the relationship between the relative directions of saccade and target motion. Each of the 12 filled circles shows 1 initial target position, and each of the 24 arrows shows the area of the visual field covered by the first 150 ms of rightward or leftward target motion at 20°/s. +, position of fixation. C and D: graphs that summarize the results for 2 monkeys. The y axes plot the eye velocity obtained by evaluating the fits to Equation 1 at 200 ms after the onset of target motion. The x axes plot the mean direction of the first saccade. Each arrow summarizes the results for 1 initial position and direction of target motion. Filled arrow and open arrow, pre- and postsaccadic eye velocity, respectively. Rightward arrows connected by dashed lines show data for rightward target motion. Leftward arrows connected by solid lines show data for leftward target motion. The 2 vertical dashed lines in each graph indicate purely upward (90°) and downward (270°) saccades, which would be orthogonal to the horizontal target motion.

Figure 8A plots presaccadic (open square) and postsaccadic (filled square) eye velocity as a function of the time from the onset of ramp target motion to the time of the measurement when the target started at 6° up and 3.75° left and moved to the right at 20°/s. This graph is almost identical to those in earlier figures and shows that there was clearly postsaccadic enhancement of horizontal smooth pursuit, even though the average saccade for this combination of step and ramp was almost purely vertical. For the trials that had no delay between the target step and ramp (squares), the horizontal saccade components averaged 0.13 ± 0.7° and the vertical saccade components averaged 5.7 ± 0.5°.

Quantitative analysis of the data from two monkeys demonstrated that postsaccadic enhancement of horizontal pursuit was largely independent of the average direction of the enhancing saccade. For each of the target steps and ramp directions illustrated in Fig. 8B, I fitted Eq. 1 to the values of pre- and postsaccadic eye velocity evoked by trials that contained different delays between the target step and the onset of the target ramp. I then evaluated each curve to obtain the value of pre- and postsaccadic eye velocity 200 ms after the onset of the ramp target motion. I also computed the average direction of the saccades evoked in the trials that had no delay between the target step and ramp. Figure 8, C and D, summarizes this analysis by plotting presaccadic (open arrows) and postsaccadic (filled arrows) eye velocity 200 ms after the onset of target motion as a function of the direction of the first saccade. Because of the geometry of the experiment, there is little overlap in the directions of the saccades evoked by the rightward and leftward target motion (plotted as appropriately directed arrows).

The analysis in Fig. 8, C and D, revealed that postsaccadic eye velocity (filled arrows) did not depend consistently on the average direction of the first saccade, although it was higher for rightward target motion (points connected by dashed lines) than for leftward target motion (points connected by solid lines) at least in one of the monkeys. In contrast, presaccadic eye velocity open arrows was consistently larger when the target moved toward the vertical meridian (vertical dashed lines) and the saccade was nearly orthogonal to the direction of target motion. I take this as evidence that postsaccadic enhancement was independent of the relative directions of the saccade and the target motion. There was less need for postsaccadic enhancement to bring eye velocity up to target velocity (i.e., presaccadic eye velocity was higher) when the saccades were orthogonal to the direction of target motion.

It is possible that the apparent preservation of postsaccadic enhancement for all directions of saccades in Fig. 8, C and D, results from the fact that most of the saccades in each average had small deviations from vertical, even though the average saccades were almost perfectly vertical in some cases. To verify that postsaccadic enhancement was preserved even in individual trials, I sorted the data according to the difference between the direction of the saccade and pure upward or downward. In Fig. 9, I plotted the presaccadic (crosses) and postsaccadic (open and filled symbols) eye velocity from individual trials, where different symbols indicate whether the direction of the saccade was within 0-1° (open circle), 1-2° (filled circle), 2-3° (open triangle), or 3-4° (filled triangle) of vertical. In each graph, I also plotted the average of the curves obtained by fitting Eq. 1 to the full data sets for presaccadic (dashed line) or postsaccadic (solid line) eye velocity, for rightward or leftward target motion.

View larger version (32K): [in this window] [in a new window]   FIG. 9. Persistence of postsaccadic enhancement of the initiation of pursuit for individual trials in which the saccade was orthogonal to the direction of target motion. Each graph plots presaccadic eye velocity (cross) or postsaccadic eye velocity (open and filled symbols) as a function of the interval between the onset of ramp target motion and the time of the measurement. Different graphs show data for leftward (C and D) or rightward (A and B) target motion and for monkey K (A and C) and monkey I (B and D). Data were sorted according to the direction of the saccade and only trials in which saccade directions were within 4° of vertical have been included. Different symbols show the angle between the direction of the saccade and vertical as shown by the key (top left in D). All data were taken from trials in which the delay between the target step and ramp was 0 or 50 ms. The solid curves show average fits of Eq. 1 to the graphs of postsaccadic eye velocity vs. latency from the onset of target motion for all initial positions that contributed data points to the graph. The dashed curves show same average fit for presaccadic eye velocity. Zero eye velocity is marked with horizontal dashed line.

Inspection of the four graphs in Fig. 9 shows that points for presaccadic eye velocity were grouped around the function for presaccadic eye velocity and were clearly below the function for postsaccadic eye velocity. The points for postsaccadic eye velocity were grouped around the average function for postsaccadic eye velocity and clearly above the function for presaccadic eye velocity. In Fig. 9, A, B, and D, however, the cluster of postsaccadic eye velocities is centered below the average function. Thus these data fall short of proving that postsaccadic enhancement of smooth pursuit is as strong when the saccade is orthogonal to the direction of target motion as it is when the saccade is in the direction of target motion. However, they show that postsaccadic enhancement of pursuit is preserved for all directions of saccades.

Effect of initial target position on pre- and postsaccadic pursuit

Lisberger and Westbrook (1985) showed that the initial eye acceleration of presaccadic pursuit depended strongly on the size of the step in a step-ramp target motion. Eye acceleration was larger for small steps than for large steps and there was a toward-away asymmetry such that target motion toward the position of fixation elicited much larger eye acceleration than did target motion away from the position of the fixation. In this section, I ask whether the same relationships appear in the enhanced postsaccadic eye velocities.

The design of these experiments was the same as that illustrated in Fig. 2C. For each combination of target step and ramp, different trials included delays of 0, 50, 100, or 150 ms between the target step and the onset of the target ramp. Target steps varied in 3° increments from 15° left to 15° right and target ramps were either rightward or leftward at 20°/s. As before, I measured the eye velocity immediately before and after the first saccade for each trial and plotted pre- and postsaccadic eye velocity as a function of the time from the onset of ramp target motion to the time of the measurement. For target motion toward the position of fixation, I included data for all initial positions. For target motion away from the position of fixation, however, I excluded data obtained with initial positions of 12 or 15°. For these initial positions, the target moved for only 100-300 ms before it had to be stopped because it had reached the edge of the oscilloscope screen. The short duration of motion, the prospect that the target would be stationary by the end of the first saccade, and the likelihood that these facts would not escape the monkey's attention made it difficult to ensure that the monkey was trying to generate smooth eye movements.

Figure 10 shows a family of graphs that illustrate the effect of initial target position on the time course of presaccadic (cross) and postsaccadic (square) eye velocity for rightward target motion at 20°/s in one monkey. For each initial target position, the postsaccadic eye velocity was large and was fitted by curves that reach asymptotic values in excess of 15°/s for almost all initial positions. As I had reported before (Lisberger and Westbrook 1985), presaccadic eye velocity varied as a function of initial target position. It was essentially the same as postsaccadic eye velocity when the target stepped 3 or 6° to the left and ramped at 20°/s to the right, but it declined as a function of initial position for both leftward and rightward positions. As before, I summarized these data by fitting Eq. 1 to the points for pre- and postsaccadic eye velocity, and I obtained a single number to quantify each response by evaluating the fit 200 ms after the onset of target motion.

View larger version (42K): [in this window] [in a new window]   FIG. 10. Postsaccadic enhancement of the initiation of pursuit for targets that stepped to different positions on the horizontal axis and moved to the right at 20°/s. Each graph shows data from 1 target step. Each point shows data from 1 trial. The y axes plot presaccadic (cross) or postsaccadic (square) eye velocity and the x axes plot the interval from the onset of target motion to the time of the measurement. Each graph includes data obtained with delays of 0, 50, 100, or 150 ms between the step and ramp of target motion. The solid curves show results of fitting Eq. 1 separately to the measurements of pre- and postsaccadic eye velocity. Numbers in the bottomleft corner of each graph give the target position after the initial target step. "Fix" indicates that there was no step of target position. The horizontal dashed lines show 0 target velocity. Data are from monkey K. From A-I, the normalized squared errors of the fits were: 5.51, 4.72, 2.67, 1.52, 2.30, 2.98, 2.18, 1.61, and 1.69, repsectively.

Figure 11, A-C, shows that the presaccadic eye velocity in our experiments depended on initial target position in exactly the same way that Lisberger and Westbrook (1985) had reported. When the analysis was done in the way described here, on trials that included early saccades (large open arrows), presaccadic eye velocity was largest when the target started close to the position of fixation. Presaccadic eye velocity declined as a function of increasingly eccentric initial positions. There also was a clear toward-away asymmetry. At any given initial position, presaccadic eye velocity was substantially larger for target motion toward versus away from the position of fixation. When the analysis was done by measuring presaccadic eye velocity 200 ms after the onset of target motion from trials in which saccades were delayed more than 300 ms after the onset of ramp target motion, I obtained almost identical results (Fig. 11, A-C, small open arrows). Saccades were delayed long enough to allow measurement of presaccadic eye velocity, however, only for target motion toward the position of fixation.

View larger version (27K): [in this window] [in a new window]   FIG. 11. Summary of the effect of the initial position of the moving target on the pre- and postsaccadic eye velocity. Top and bottom rows of graphs show data for pre- and postsaccadic eye velocity, respectively. In each graph, the large arrows plot the eye velocity obtained by evaluating the fits of Eq. 1 at 200 ms after the onset of target motion, as a function of the initial target position created by the target step. Thus each connected set of points summarizes the analysis of a full set of graphs like those in Fig. 8. Top row: the small arrows plot the eye velocity 200 ms after the onset of target motion in trials for which the saccade latency was longer than 200 ms. For the most part, saccades were delayed this long only when the initial target position was 3-12° eccentric and target motion was toward the position of fixation. Rightward and leftward arrows show the results for rightward and leftward target ramps, respectively.

Figure 11, D-F, demonstrates that postsaccadic eye velocity evinced little or no relationship to initial target position. Comparison of the two data points at each initial target position reveals a few examples where postsaccadic eye velocity is larger for target motion toward versus away from the position of fixation. However, there is no convincing evidence of the pervasive toward-away asymmetry that has been omnipresent in studies of presaccadic pursuit.

	DISCUSSION

Abstract Introduction Methods Results Discussion References

I have demonstrated that there is strong postsaccadic enhancement of pursuit eye movements. The postsaccadic enhancement can be seen most easily when the target motion takes the target away from the position of fixation because presaccadic pursuit is especially weak under these conditions. However, postsaccadic enhancement of pursuit occurs when the saccade is in the same or the opposite direction from target motion and even when the direction of the saccade is orthogonal to the direction of the enhanced pursuit. This makes it difficult to explain postsaccadic enhancement of pursuit as a mechanical effect mediated in the orbit. Unlike other forms of postsaccadic enhancement of smooth eye movements (e.g., Busettini et al. 1996; Kawano and Miles 1986), postsaccadic enhancement of pursuit is provoked by the execution of the saccade per se and not by the rapid motion of retinal images during the saccade. Thus it seems likely to be mediated by a nonvisual input to the visual-motor pathways for pursuit, presumably from motor signals related to the planning or execution of a saccade.

What is postsaccadic enhancement of pursuit?

I think that postsaccadic enhancement of the initiation of pursuit is related closely to a pursuit "switch" that seems to regulate the gain of visuo-motor processing for pursuit and that must be "on" to allow effective tracking. Two of our prior studies provide the most compelling evidence in favor of a pursuit switch. Goldreich et al. (1992) showed that the gain of smooth eye velocity for high-frequency target vibration is much higher if the vibration is imposed while a monkey is tracking a moving target than if it is imposed while the monkey is fixating a stationary target. Schwartz and Lisberger (1994) showed that the size of the pursuit response to a brief perturbation of target velocity depends on the conditions at the time of the perturbation. The response is small when the monkey is fixating and grows as a function of ongoing target velocity.

Additional evidence for a pursuit switch is provided by two examples of deficits in smooth pursuit tracking without deficits in motion processing. In one example, I reported the pursuit eye movements of a monkey that had very weak initiation of pursuit for upward target motion but that could respond normally to upward image motion if it was presented during downward target motion and that also could generate brisk upward smooth eye movements when the moving stimulus was a large texture (Grasse and Lisberger 1992). In the second example, monkeys with early-onset artificial strabismus have very weak initiation of pursuit for temporally directed target motion with monocular viewing, but they respond normally to temporally directed image motion during nasally directed pursuit (Kiorpes et al. 1996). In both of these cases, the interpretation was that these animals had normal visual motion processing in all directions but could not turn on the pursuit switch for selected directions. In the latter example, normal visual motion processing was confirmed by single unit recordings showing a normal distribution of preferred velocities and directions among units in visual area MT, even with monocular viewing.

Previous experiments have revealed the paradox that target motion away from the position of fixation is a very poor stimulus for presaccadic pursuit but that eye velocity can be close to target velocity immediately after the saccade. It generally has been assumed that the smooth eye movements immediately after a saccade velocity must be driven by visual inputs that were present before the saccade (e.g., Groh et al. 1997; Newsome et al. 1985). If, as seems inescapable, this assumption is true, then why are the same visual inputs rather ineffective before the saccade? I propose the following resolution of this paradox. Target motion away from the position of fixation provides strong visual motion inputs but is a poor stimulus for activation of the pursuit switch and a strong stimulus for saccades. As a result, there is an early saccade and presaccadic pursuit is weak in spite of the presence of a strong visual motion signal. For this configuration of target motion, the pursuit switch is activated not by target motion but rather by signals related to the planning or execution of the saccade to the moving target.

Two parallel processes in the initiation of pursuit

The hypothesis of a pursuit switch implies that there are two separate processes that operate in parallel during the initiation of pursuit. It is plausible to assume that these two processes could have dissimilar motion processing and that they might be implemented in different neural pathways. One process involves sensing visual motion and computing the correct eye acceleration to bring eye velocity up to target velocity. The second process also senses visual motion but instead decides whether the pursuit switch should be activated and if so, how strongly. One can think of the pursuit switch as a volume control that is potentially under the control of a number of different neural processes. Even very strong image motion does not cause large velocities of smooth eye movement if the pursuit switch has not been activated. However, neural signals related to saccades or the decision to makes a saccade seem to have immediate access to the pursuit switch and to be able to turn it on fully and quickly.

The demonstration that postsaccadic enhancement largely obviates the effects of initial target position on the presaccadic initiation of pursuit may require some revision of the interpretations based on analysis of presaccadic pursuit. For example, Lisberger and Westbrook (1985) assumed that presaccadic pursuit was a probe for the motion processing in the visuo-motor pathways that guide pursuit. They argued that the weak initiation of pursuit for eccentric targets reflected a paucity of inputs from eccentric parts of the visual field to the pursuit system. Our data on the effect of initial target position on presaccadic smooth eye velocity agree with those of Lisberger and Westbrook (1985). If we consider postsaccadic eye velocity, however, then motion across eccentric visual field (15°) has almost as strong an effect on the initiation of pursuit as does motion across central visual field. It is not possible to explain the uniformly large postsaccadic smooth eye velocities as consequences of the closed-loop, negative-feedback architecture of the pursuit system. As Newsome et al. (1985) pointed out, the immediate postsaccadic smooth eye velocity must result from visual inputs that were present before the saccade. Further, our curve fits are based heavily on data that were obtained within 200 ms of the onset of target motion and, therefore, within the traditional "open-loop" interval in the initiation of pursuit (Lisberger and Westbrook 1985).

To account for the strong effect of initial target position on pre-saccadic eye velocity at the initiation of pursuit, I propose that the initial position of the moving target is an important factor in the control of the pursuit switch. If this were the case, then targets moving across different positions in the visual field could cause equally strong visual response that could be expressed by uniformly large eye velocities after a saccade to the target had fully activated the pursuit switch. Thus in contrast to the interpretation of Lisberger and Westbrook (1985), the relationship between presaccadic eye velocity and initial target position need not reflect anything about the relative strength of visual motion signals from different parts of the visual field. I also suggest that the pronounced toward-away asymmetry in the initiation of pursuit cannot be attributed to a toward-away bias in the inputs to the visual-motor pathways that guide pursuit. Instead, it appears that targets moving away from the position of fixation are simply ineffective at activating the pursuit switch. In future research, it will be important to ascertain whether a given stimulus causes weak presaccadic pursuit because it provides weak visual motion signals or because it is not suitable for activating the pursuit switch even in the presence of strong visual motion signals.

Postsaccadic enhancement as one way the brain selects pursuit targets

One of the major defining features of pursuit eye movements is that they are generated by a voluntary motor system. The pursuit system can track in the absence of target motion if target motion is anticipated (Kowler and Steinman 1981). Even if a moving target is present, the pursuit system need not elect to track it. If two targets are present, the pursuit system can choose between them. In two previous papers, for example, Ferrera and Lisberger (1995, 1997) have demonstrated that monkeys can be trained to select one of two moving targets based on a colored cue.

I now propose that the decision to make a saccadic eye movement to a target provides a powerful way to tap into the ability of the pursuit system to select targets. Based on the data in the present paper, it is plausible to think that the pursuit system gets activated by any target whose images are soon going to be placed on the fovea. By this criterion, targets moving toward the position of fixation from 6° eccentric and targets that are about to be the target of the saccade are both excellent stimuli for activating the pursuit switch. Under natural conditions, a tight connection between the decision to make a saccadic eye movement and the selection of a pursuit target seems like an excellent design principle. One major advantage of this approach is that the neural circuits for selecting from multiple targets need not be created separately for both saccades and pursuit. If these selection circuits decide that a target will be acquired by a saccade, then the motion signals from that target automatically acquire large weights as stimuli for pursuit.

A linkage between the selection of a target for pursuit and saccades could be implemented in two ways. Either a single set of decision circuits could control saccades and pursuit in parallel or the decision circuits could select the target for a saccade and then rely on the motor signals from the saccade to select a pursuit target. My data don't distinguish between these two possibilities. However, we can learn something about the frame of reference of the target selection mechanism for pursuit from the fact that postsaccadic enhancement occurs even when pursuit is orthogonal to the direction of the enhancing saccade. Specifically, it seems that the selection of a target for a saccade focuses the attention of the pursuit system on the visual motion inputs from a the part of the visual field that contains the saccade target rather than on visual motion signals in the same direction as the saccade.

	ACKNOWLEDGEMENTS

  I am grateful to S. Tokiyama for careful data analysis. I thank F. Miles, whose earlier publications provided the inspiration for these experiments, M. Kahlon for many insightful discussions, and the other members of my laboratory for helpful comments on an earlier version of the manuscript.

  This research was supported by National Eye Institute Grant EY-03878.

	FOOTNOTES

  Address for reprint requests: Dept. of Physiology, 513 Parnassus Ave., Room 762-S, UCSF, Box 0444, San Francisco, CA 94143.

  Received 10 June 1997; accepted in final form 3 December 1997.

	REFERENCES

Abstract Introduction Methods Results Discussion References

• BUSETTINI, C., MILES, F. A., KRAUZLIS, R. J. Short-latency disparity vergence responses and their dependence on a prior saccadic eye movement. J. Neurophysiol. 75: 1392-1410, 1996.[Abstract/Free Full Text]
• FERRERA, V. P., LISBERGER, S. G. Attention and target selection for smooth pursuit eye movements. J. Neurosci. 15: 7472-7484, 1995.[Abstract]
• FERRERA, V. P., LISBERGER, S. G. The effect of a moving distractor on the initiation of smooth pursuit eye movements. Vis. Neurosci. 14: 323-338, 1997.[Medline]
• GOLDREICH, D., KRAUZLIS, R. J., LISBERGER, S. G. Effect of changing feedback delay on spontaneous oscillations in smooth pursuit eye movements of monkeys. J. Neurophysiol. 67: 625-638, 1992.[Abstract/Free Full Text]
• GRASSE, K. L., LISBERGER, S. G. Analysis of a naturally occurring asymmetry in vertical smooth pursuit eye movements in a monkey. J. Neurophysiol. 67: 164-179, 1992.[Abstract/Free Full Text]
• GROH, J. M., BORN, R. T., NEWSOME, W. T. How is a sensory map read out? Effects of microstimulation in visual area MT on saccades and smooth pursuit eye movements J. Neurosci. 17: 4312-4330, 1997.[Abstract/Free Full Text]
• JUDGE, S. J., RICHMOND, B. J., CHU, F. C. Implantation of magnetic search coils for measurement of eye position: an improved method. Vision Res. 20: 535-538, 1980.[Medline]
• KAWANO, K., MILES, F. A. Short-latency ocular following responses of monkey. II. Dependence on a prior saccadic eye movement. J. Neurophysiol. 56: 1355-1380, 1986.[Abstract/Free Full Text]
• KIORPES, L., WALTON, P. J., O'KEEFE, L. P., MOVSHON, J. A., LISBERGER, S. G. Effects of early-onset artificial strabismus on pursuit eye movements and on neuronal responses in area MT of macaque monkeys. J. Neurosci. 16: 6537-6553, 1996.[Abstract/Free Full Text]
• KOWLER, E. The role of visual and cognitive processes in the control of eye movement. In: Eye Movements and Their Role in Visual and Cognitive Processes, edited by and E. Kowler Amsterdam: Elsevier, 1990, p. 1-70.
• KOWLER, E., STEINMAN, R. M. The effect of expectations on slow oculomotor control. III. Guessing unpredictable target displacement. Vision Res. 18: 603-605, 1981.
• LISBERGER, S. G., WESTBROOK, L. E. Properties of visual inputs that initiate horizontal smooth pursuit eye movements in monkeys. J. Neurosci. 5: 1662-1673, 1985.[Abstract]
• NEWSOME, W. T., WURTZ, R. H., DÜRSTELER, M. R., MIKAMI, A. Deficits in visual motion processing following ibotenic acid lesions of the middle temporal visual area of the macaque monkey. J. Neurosci. 5: 825-840, 1985.[Abstract]
• ROBINSON, D. A. The mechanics of human smooth pursuit eye movement. J. Physiol. (Lond.) 180: 569-591, 1965.[Free Full Text]
• SCHWARTZ, J. D., LISBERGER, S. G. Initial tracking conditions modulate the gain of visuo-motor transmission for smooth pursuit eye movements in monkeys. Vis. Neurosci. 11: 411-424, 1994.[Medline]

0022-3077/98 $5.00 Copyright ©1998 The American Physiological Society