Interactions between Saccades and Smooth Pursuit Eye Movements in Marmosets

Animals use a combination of eye movements to track moving objects. These different eye movements need to be coordinated for successful tracking, requiring interactions between the systems involved. Here, we study the interaction between the saccadic and smooth pursuit eye movement systems in marmosets. Using a single-target pursuit task, we show that saccades cause an enhancement in pursuit following a saccade. Using a two-target pursuit task, we show that this enhancement in pursuit is selective toward the motion of the target selected by the saccade, irrespective of any biases in pursuit prior to the saccade. These experiments highlight the similarities in the functioning of saccadic and smooth pursuit eye movement systems across primates.


Introduction
Primates use a combination of eye movements to visually track objects in our environment (Rashbass, 1961).Saccadic eye movements bring objects of interest onto the fovea whereas smooth pursuit eye movements stabilize moving objects onto the retina.These two eye movements are sensitive to distinct aspects of object information: saccades are primarily sensitive to object position, and pursuit eye movements are primarily sensitive to object motion (Rashbass, 1961).While these two classes of eye movements are distinct, the retinal input and scene interpretation driving these movements are shared (Krauzlis et al., 1997;Gardner andLisberger, 2001, 2002;Adler et al., 2002;Krauzlis and Dill, 2002;Liston andKrauzlis, 2003, 2005;Carello and Krauzlis, 2004;Krauzlis, 2004;Schoppik and Lisberger, 2006;Case and Ferrera, 2007).Achieving the common goal of selecting and tracking visual objects thus requires the coordination of both saccades and smooth pursuit eye movements.
Marmosets are emerging as a primate model in vision, offering specific advantages in studying the neural circuit functions underlying visual functions.Their brain shows a similar compartmentalization of areas as in other primates, including areas associated with visual processing and eye movement behaviors.The cortical areas, in particular, are more accessible due to the lissencephalic marmoset cortex.This feature makes marmosets suitable for mesoscale optical and electrophysiological circuit investigations (Pattadkal et al., 2023b).While the cortical visual and eye movement areas are similar to other primates, it is not known whether their eye movement behaviors share similar characteristics as in macaques or humans.We, and others, have already demonstrated that their saccadic behaviors are comparable to humans and macaques with small differences in saccade rates and sizes (Mitchell et al., 2014;Samonds et al., 2024).They also exhibit certain motion-dependent eye movements such as ocular following with similar features to those of macaques and humans (Yip et al., 2023;Pattadkal et al., 2023a).In this study, we examine the interaction of saccade and smooth pursuit behavior in marmosets and compare the influence of saccades on pursuit to previous studies from macaques and humans.
Here we examined how saccades alter pursuit eye movements in situations in which marmosets track either single or multiple targets.While the primary drive to pursuit eye movements is target motion, it has long been recognized that pursuit is susceptible to extraretinal factors (Schwartz and Lisberger, 1994;Ferrera and Lisberger, 1995).These factors, including a selection of a target, influence the direction of pursuit eye movements (Stone et al., 1996;Beutter and Stone, 1998) and the degree to which the pursuit system is engaged.Saccadic eye movements to a target, potentially reflecting target selection, lead to an increase in the amplitude of smooth pursuit in humans (Robinson, 1965;Luebke and Robinson, 1988) and macaques (Lisberger, 1998).This post-saccadic enhancement has been interpreted as a reflection of the engagement of the pursuit system (Schwartz and Lisberger, 1994;Churchland and Lisberger, 2002) when a target has been selected.Evidence for this engagement also exists in marmosets: responses to a target perturbation during pursuit are larger than the same target perturbation from fixation (Mitchell et al., 2015).It is unclear, however, if this engagement is related to the saccadic system, as these perturbations were made during steady state pursuit.
The visual world is often composed of multiple moving objects.To selectively track a single moving object, animals generally saccade to an object, to reduce the positional error and then pursue that object to reduce the velocity error.As such, there is a need for coordination between the two eye movements to ensure that the saccade and pursuit match the position and velocity of the selected target.Previous studies in humans (Krauzlis et al., 1999) and macaques (Gardner and Lisberger, 2001) have shown that target selection decisions for saccade and pursuit are indeed linked, but it remains unclear whether these systems are working in parallel or serial.Here we demonstrate that, as in macaques and humans, saccades influence the gain of pursuit eye movements.Pursuit is enhanced toward motion of the saccade target.We also find, however, that smooth eye movements in marmosets exhibit information about target selection prior to saccadic eye movements, indicating that choice about pursuing an object does not require saccadic eye movements.We show that saccadic selection carries more influence on altering pursuit selection while pre-saccadic pursuit selection only weakly relates with saccade selection and does not affect post-saccadic pursuit enhancement.

Materials and Methods
Two male common marmosets were used for this study.The animals were surgically implanted with a titanium headpost, attached to the skull using Metabond (Parkell).Post recovery, they were trained to be head-fixed and perform fixation and pursuit (Mitchell et al., 2015;Pattadkal et al., 2023a).To motivate animals for behavioral tasks, they were on food control with weights maintained at 85-90% of baseline weight.Baseline weight was calculated as average weight during a 4 week period each year when the animal is on ad libitum food and water and no behavioral training occurs.All procedures conformed to National Institutes of Health guidelines and the guidelines of The University of Texas at Austin Institutional Animal Care and Use Committee.
We used the EyeLink 1000 Plus system to track the right eye (SR Research).The data was collected in pupil-CR mode at 1 kHz rate.Stimuli were displayed on FlexScan T761 50 cm (19 inch) Class Color Display cathode ray tube monitor with a refresh rate of 85 Hz.The distance between monitor and the animal is 50 cm with a field of view of 35°by 28°.Stimuli were shown on a gray background with mean luminance 31.5 cd/m 2 .The maestro software was used to control the behavioral tasks and collect data (https://sites.google.com/a/srscicomp.com/maestro/).
Tasks started with animals fixating at a target presented at the center of the screen for 500-1,000 and 200-500 ms grace period.Marmoset faces or small dots were used as targets.For single-target tasks, the target then stepped either in the direction (positive step) or in the opposite direction (negative step) of upcoming target motion.The target then moved at a constant speed for 700-800 ms.For a successful trial, animals had to follow the moving target, remaining within a fixation window of 2°.For Animal 1, target moved at 5°/s and step size was +1 or −1°.Target directions were either rightward (0), upward (90), or leftward (180).For Animal 2, the target moved at 8°/s and step size was −1.6°.Target directions were either rightward (0), upward (90), leftward (180), or downward (270).For two-target tasks, after the initial fixation, two targets appeared and made a step jump.Both targets then moved in their respective direction for up to 800 ms.After a variable period, the target not chosen by the saccade was turned off.For successful trial, animals had to follow the moving target, remaining within a fixation window of 2°.For Animal 1, targets moved at 5°/s and step size was +1 or −1°.Target directions were rightward (0) and upward (90).For Animal 2, targets moved at 8°/s and step size was −1.6°.Target directions were either rightward (0) and downward (270) or leftward (180) and upward (90).Single-and two-target stimulus conditions were interleaved.Successful trials were rewarded with marshmallow juice.
Saccade onset and offsets were manually marked using JMWork (https://sites.google.com/a/srscicomp.com/maestro/data-analysis/jmwork) as in Mitchell et al. (2015).Eye velocity was computed from the eye position trace which was collected at 1 kHz and median filtered over 25 ms.The filtered position trace was then differentiated at 25 and 50 Hz, and the mean was used as velocity estimate (Mitchell et al., 2015;Pattadkal et al., 2023a).We computed pursuit latency using single-target pursuit trials, separately for each animal, for each step size and for each direction (Table 1).To estimate pursuit latency, we examined trials that did not have a saccade within 200 ms of the start of target motion.Each trial position raw trace in the direction of target motion was smoothed using a median filter of length 5 and the average position in the first 50 ms after motion start was subtracted from this trace.The average position in a 200 ms interval from start of target motion was fit with a threshold linear function with the following form: here, e p is the average eye position, g is a gain factor, t is time of the trace being 0 at motion onset, l is the pursuit latency, and c is the offset.The latency measurement was estimated using a bootstrap procedure, resampling the eye position traces 1,000 times (Sokal and Rohlf, 1995).The mean of the generated latency distribution was used as the mean pursuit latency for that condition.Mean latencies averaged across directions for Animal 1 positive stepped target motion was 191 ms, for Animal 1 negative stepped target motion was 190 ms, and for Animal 2 negative stepped target motion was 177 ms.Pre-and post-saccadic eye velocities were computed per trial from 30 ms windows before and after the saccade.The pre-saccadic period ended 25 ms before the actual saccade onset (54-25 ms before saccade onset).The post-saccadic period started 25 ms after the actual saccade offset (25-54 ms after saccade offset).To ensure that pursuit is occurring in the pre-saccadic analysis window, we have only considered trials in which the saccade latency is higher than pursuit latency +55 ms.For comparison with velocity during the fixation period, average velocity during fixation was also computed in a 30 ms window before motion onset.Additionally, trials were only included in further analysis if the amplitude of pre-and post-saccadic pursuit exceeded 0.5°/s.Post-saccadic eye positions were computed per trial from a 10 ms window after the saccade, starting 25 ms after the actual saccade offset (25-34 ms after saccade offset).
Saccade and pursuit weights were computed as in Gardner and Lisberger (2001) with some modifications.Saccade weights were computed by using the following equation: where w is the saccade weight.T 1 and T 2 represent the target position unit vectors for the two targets at time of the saccade.E is the unit eye position vector for post-saccadic eye position.
Pursuit weights were computed by using the following equation: where w is the pursuit weight.T 1 and T 2 represent the average unit pursuit velocity vectors during single-target pursuit of target 1 or 2, respectively.E is the unit eye velocity vector from the trial.For pre-saccadic pursuit weight, T 1 , T 2 , and E are from pre-saccadic eye velocities.For post-saccadic pursuit weight, T 1 , T 2 , and E are from post-saccadic eye velocities.
To compute the relation between pursuit and saccade weights, a linear fit of the following form was applied: where y is the pursuit weight, x is the saccade weight, m is the slope of the fit, and c is the y-intercept.
Choice probability was estimated by measuring the area under the ROC curve (Green and Swets, 1974).Briefly, pre-or post-saccadic pursuit weights were divided into two distributions based on the saccade choice for each trial.The separability of the two distributions was measured by varying the criterion.The hit rate and false alarms obtained for each criterion were used to make the ROC curve.To compute the error bars over choice probability values shown in Figure 5, we sampled the weights with replacement 1,000 times and used the resulting choice probability distribution to compute the 95% confidence intervals.
Code accessibility.JMWork, the program we use to mark eye movements is freely available at https://sites.google.com/a/srscicomp.com/maestro/data-analysis/jmwork.

Results
Post-saccadic enhancement of single-target pursuit We first examined the relationship between saccadic and pursuit eye movements when animals tracked a single target using the step-ramp pursuit task (Fig. 1).We are interested in how the presence of a saccadic eye movement at the initiation of pursuit alters the pursuit eye velocity that immediately follows the saccade.If saccadic eye movements reflect an engagement of the pursuit system, we expect that post-saccadic pursuit will have a higher gain than pre-saccadic pursuit and a higher gain than pursuit trials in which no saccade had been made (Lisberger, 1998).Such a comparison requires variability in the latency of the initial saccade during the pursuit trials, which the pursuit behavior provides naturally or could be enhanced by changing the initial position of the moving target.The task began with the animal fixating at a target at the center of the screen after which a target was stepped either in the direction (positive step) or opposite to the direction (negative step) of target motion and then the target moved (Fig. 1A).We used both step configurations for Animal 1 and only negative steps for Animal 2. Both animals successfully tracked the moving target along different directions using a combination of saccades and pursuit (Fig. 1B-D).Pre-saccadic pursuit is generally slower, whereas once the animal makes a saccade to the moving target, the eye velocity increases to match target velocity (Fig. 1B-D).
To quantify the effect of saccades on the initiation of pursuit eye movements, we examined the eye velocity signals before and after the first saccade in each pursuit trial.We used an EyeLink 1000 video eye tracker to measure eye position, and we observe a systematic perturbation in the eye position traces near the edge of the saccadic eye movements.The eye appears to stop or reverses direction at the edge of the saccade, as others have also noted (Kimmel et al., 2012).These distortions, which may be due to changes in the shape of the pupil during the saccade, could potentially disrupt our measurements of pre-and post-saccadic pursuit.We therefore did not examine pursuit within 25 ms of saccadic eye movements.To ensure that pursuit is occurring in the pre-saccadic pursuit analysis window, we only considered trials with saccade latencies exceeding pursuit latencies plus the length of the analysis and buffer window (see Materials and Methods).We also compared the distribution of pre-saccadic pursuit velocities to velocities during fixation period using t test, to show that eye is indeed moving in the pre-saccadic pursuit window (Table 1).
To measure the effect of the saccade on pursuit gain, we first aligned pursuit trials on the timing of the first saccade and averaged the measured eye velocity that preceded and followed the saccade.On average eye velocity increased in target direction following saccade relative to the eye velocity in the same direction preceding the saccade.This effect existed across animals, across step sizes, across most target motion directions, and across target velocities (Fig. 2A-C, Table 1).The amount of increase in eye velocity following saccade was not found to be related with saccade duration (R 2 < 0.05 for all single-target motion conditions across both animals) or with saccade amplitude (R 2 < 0.05 for all single-target motion conditions for Animal 1, and three out of four single-target motion conditions for Animal 2).This increase in eye velocity could be due to the natural increase in gain that comes at the initiation of pursuit (Schwartz and Lisberger, 1994).That is, while we observe an increase related to the saccade, we already expect an increase to occur because the pursuit system has not yet matched the velocity of the target motion.To distinguish between the effects of time and those of the saccade itself on pursuit velocity, we binned trials based on the time of measurement around which the velocity window was centered and compared the pre-and post-saccadic amplitudes for measurements made at the same time (Fig. 2A-C, third column).Velocities generally show some increase with time for pre-and post-saccadic pursuit.But, the post-saccadic pursuit velocity exceeded pre-saccadic pursuit velocity for velocities measured at same time points.Therefore, the rise in pursuit velocity in post-saccadic period is not due to a simple progression in time during the trial, but an effect of the saccade itself.As in other primates (Lisberger, 1998), marmoset pursuit gain is increased by the occurrence of a saccadic eye movement.

Pursuit saccade interaction in a two-target task
We have demonstrated that for a single target, pursuit gain is linked to saccadic eye movements.In the natural world, however, there may be multiple moving objects moving in distinct directions, and animals need to select which target to track.Target selection may engage a combination of saccades and pursuit to foveate the selected target and to stabilize the target motion on the retina.In macaques, a complex interplay between saccades and pursuit has been revealed by employing a task in which two targets are presented and moved, and the animal is given the option to track either target for a reward (Gardner and Lisberger, 2001).As in the single-target case, it was found that an increase in pursuit of a target is linked to the saccadic eye movement, but also the direction of the pursuit changed following a saccade.Prior to an initial saccade, pursuit eye movements moved in a direction intermediate between the targets whereas after the saccade pursuit moved in the direction of the selected target.
We examined whether there is a similar linkage between saccades and pursuit in marmosets in this two-target task.Animals began with a fixation at the center of screen.Two targets then appeared and moved in different directions at the same speed following a target step.The animal was free to track either target and the other target was extinguished following a variable period (Gardner and Lisberger, 2001; Fig. 3A).To determine the linkage between the saccadic selection and smooth pursuit, we compared the pre-saccadic and post-saccadic pursuit velocity.If saccadic selection is linked to pursuit selection, then we would expect that eye movements following the saccade would track the velocity of the saccadic target and not of the nonselected target.Indeed, we find that post-saccadic pursuit velocity increases as in the single-target case, and this increase is toward the target selected by the saccade (Fig. 3B-E).
To quantify the evolution of choice during pursuit, we used weights to assign a degree of selectivity in pursuit and of the saccade.Saccade weights are computed by measuring proximity of saccade end point to either of the targets, 0.5 weight being equidistant from both targets and weight of 0 or 1 indicating saccade to either of the targets.Animals used saccades to land closer to one of the targets (see spread of eye position before and after a saccade in Fig. 4A).Pursuit weights were computed in a similar fashion.If the two-target pursuit were the average of two single-target pursuits, the pursuit weight would be close to 0.5, whereas a bias toward pursuit of either target shifts the weight toward 0 or 1.For each trial, we computed the pre-saccadic pursuit weight, the saccade weight, and the post-saccadic weight to determine the interaction between saccades and pursuit.

Does saccadic selection affect pursuit selection?
We first tested if saccadic selection influenced pursuit selection.A link between the two would cause post-saccadic pursuit weight to be more related to saccade weight than pre-saccadic pursuit.To quantify this, we measured the slope of the relationship between pre-and post-saccadic pursuit weights and saccade weights.We found that the post- saccadic pursuit weight was more strongly related to the saccade weight than the pre-saccadic pursuit weight (Fig. 4B, Table 2).This analysis indicates that post-saccadic pursuit is stronger toward the target that was selected by the saccadic eye movement.
An alternative way to demonstrate the linkage between saccades and pursuit is to measure the choice probability between the saccadic eye movement and the post-saccadic pursuit.This reflects the degree to which pursuit is predictive of the saccade choice.We first measured how well we could detect the saccade choice using the pursuit weights for preand post-saccadic pursuit using ROC curves.The post-saccadic pursuit ROC curve is quite convex, reflecting a strong match between the saccadic and pursuit systems following saccade.From this ROC analysis, we measured the area under the curve or choice probability.For both animals, we found that choice probability systematically increases for postsaccadic pursuit (Fig. 4C).Second column in all rows is data for Animal 1, negative step, targets are moving along 0 and 90°(n = 97).Third column in all rows is data for Animal 2, negative step, targets are moving along 180 and 90°(n = 301).Fourth column in all rows is data for Animal 2, negative step, targets are moving along 0 and 270°(n = 198).
Is selection absent in pursuit prior to the saccade?
We have demonstrated that pursuit gain increases following saccades and that the direction of pursuit matches the saccade choice post-saccade.But we also find evidence that the pre-saccadic pursuit contains information about the target choice.There are three analyses that indicate that pre-saccadic pursuit exhibits some selection.First, the relationship between the pre-saccadic weight and the saccadic weight is positive (Fig. 4B).This indicates that the pursuit is already biased toward one target.Second, in a similar fashion, the choice probability between pre-saccadic pursuit and the saccade is significantly above 0.5, indicating that there is information in the pursuit eye movements that predicts which target that will be selected.Finally, the distribution of weights reveals that pre-saccadic pursuit also exhibits some selection (Fig. 5A).If the pre-saccadic pursuit were performing a vector average computation, those weights would lie near 0.5.But instead, the pre-saccadic pursuit weights form a bimodal distribution suggesting that the pursuit system has already selected a target.The relatively weak relationship between the pre-saccadic pursuit and saccade weights indicates that it is not uncommon for the saccades to select a different target than the pursuit target.This can be observed in some trials in which saccadic choice did not match the pre-saccadic pursuit choice (Fig. 3C,E).To study how far back in time presaccadic pursuit is indicative of upcoming saccade choice, we computed choice probabilities at different intervals surrounding the saccade (Fig. 5B).The choice probability increased slightly in the interval leading up to the saccade, but the largest change is observed following the saccade.Does the saccadic gain enhancement depend on a match to pre-saccade pursuit choice?
Because it is apparent that for some trials the target selected by the pursuit system and the saccadic system are different, we wondered whether changing the selected target would impact the resulting pursuit.If increased gain toward  the selected target was affected by a lack of match between saccade and pre-pursuit choice, it would be reflected in the post-saccadic pursuit velocity.To test this idea, we compared post-saccadic pursuit velocities within each saccade choice.For each saccade choice, we divided the post-saccadic velocities into two groups based on whether the saccade choice was matched or mismatched to the pre-saccadic pursuit choice.We then compared these two groups of postsaccadic velocities (Fig. 6).We do not observe any differences in post-saccadic velocity in the matched versus unmatched groups (KS test, comparisons were made in cases with 30 or more trials per subgroup).This indicates that target selective gain enhancement post-pursuit is not dependent on a match with pre-pursuit choice.probabilities in marmosets were higher than 0.5, indicating that there is already some choice information present in the pre-saccadic pursuit.One reason for this difference may be the differences in the velocities used (20°/s in macaque study vs 5 or 8°/s in marmosets).The differences in target speeds may affect the saccade latencies (Gellman and Carl, 1991).The mean saccade latency in the macaque study was 224 ms ± 55 whereas mean saccade latency in the two-target task in our study is 391 ms ± 10 7 SD (Animal 1, positive step), 488 ms ± 98 SD (Animal 1, negative step), 433 ms ± 104 SD (Animal 2, negative step).Smaller speeds can also delay pursuit onset though (Ferrera and Lisberger, 1995).The later saccade latency in the present study may give time for the pursuit system to begin to exhibit signatures of target selection.
It seems efficient that the target selection decisions for the saccade system and pursuit system should be shared as they both have a common goal to view a target (Orban De Xivry and Lefèvre, 2007).While the selection may itself be shared between the pursuit and saccade systems (Gardner andLisberger, 2001, 2002;Liston andKrauzlis, 2003, 2005), which of the two systems reflects this selection first varies.Some studies (Gardner andLisberger, 2001, 2002;Schoppik and Lisberger, 2006) argue for a serial linkage between saccadic and pursuit target selection with saccades (either natural or evoked using microstimulation in FEF or SC) capable of triggering selection in pursuit.Other studies (Liston andKrauzlis, 2003, 2005;Case and Ferrera, 2007) show that pursuit preceding saccade can have a different selected target than the saccadic target, though pursuit changes its target to match the saccadic selection around time of saccade.This shift in pursuit can occur prior to the saccade itself with pursuit being able to predict the saccadic target at least 60 ms before the saccade (Liston and Krauzlis, 2003).Coordination dynamics between smooth pursuit and saccade can also depend on the task structure (Erkelens, 2006).The choice of target may be conveyed in parallel to the saccade and pursuit system but the execution of the following saccade and pursuit may occur independently with the time depending on the nature of the task.It is also important to note that a link between saccade and pursuit does not imply that saccades are necessary for pursuit target selection (Ferrera and Lisberger, 1995;Lisberger and Ferrera, 1997;Krauzlis et al., 1999;Garbutt and Lisberger, 2006).However, when both occur in concert, they track the same object.Our results show such pre-saccadic target selection in marmoset pursuit (target selection arrow toward pursuit system in Fig. 7).The presaccadic pursuit selection only weakly affects the saccadic target selection though (see Table 2 pre-vs post-saccadic relation), indicating independent target selection in saccades (target selection arrow toward saccadic system and weak arrow from pursuit selection to saccade selection in Fig. 7).The pre-saccadic pursuit choice also has no influence on postsaccadic pursuit enhancement.While pursuit selection may not always result in same saccade selection, selection by saccades alters or determines the pursuit target to match (strong arrow from saccadic selection to pursuit selection in Fig. 7).This may be because saccades are the primary means of directing gaze to overtly attend targets.

Significance of pursuit saccade interaction
Post-saccadic enhancement is known to occur for a variety of eye movements like smooth pursuit in macaques (Lisberger, 1998), ocular following (Kawano and Miles, 1986), disparity vergence (Busettini et al., 1996), tracking responses in marmosets (Coop et al., 2023).Unlike other eye movements however, post-saccadic enhancement in smooth pursuit is not because of image motion generated by a saccade, but due to the actual saccade.This was shown to be the case by Lisberger (1998) where he observed no enhancement in pursuit upon simulating a saccade by imposing a similar image motion.Pursuit enhancement following a saccade is thought to be a consequence of turning on of the pursuit gain.This pursuit gain is low during visual fixation (Goldreich et al., 1992;Schwartz and Lisberger, 1994) and when switched on, allows for effective tracking of moving target.
Our results demonstrate the similarities in functioning and coordination of eye movement systems between marmosets and other primates.The accessibility of the underlying neural elements in marmosets (Mehta et al., 2019;Pattadkal et al., 2023b) will allow for a circuit level dissection of this cognitive process.

Figure 1 .
Figure 1.Single-target pursuit task examples.A, The scheme for single-target step-ramp pursuit task.Animals begin by fixating at a target at center of the screen.The target then makes either a positive or negative step and starts to move.B, C, Example pursuit traces from animals.Top row is showing position trace in the direction of target movement.Bottom row is showing velocity trace in the direction of target movement.Black lines indicate the eye traces.Colored blue and red lines indicate the target motion, along vertical (blue) or horizontal (red) direction.Inset schematics show which direction the target movement is in for each example.Shaded regions indicate the 30 ms windows used to compute pre-saccadic (gray) and post-saccadic (green) pursuit velocity.B is a trace from Animal 1, for positive step target moving upward at 5°/s.C is a trace from Animal 1, for negative step target moving rightward at 5°/s.D is a trace from Animal 2, for negative step target moving leftward at 8°/s.

Figure 2 .
Figure 2. Post-saccadic enhancement of single-target pursuit.A, Analysis from Animal 1 for positive step-ramp trials (n = 169).First panel shows average eye velocity before (gray) and after (green) the saccade for rightward target motion.Dotted lines indicate bootstrapped confidence intervals.Inset shows the target motion.Second panel shows the distribution of pre and post-saccadic eye velocity for all rightward single-target motion trials.Green indicates post-saccadic pursuit and gray indicates pre-saccadic pursuit.Third panel compares pre (gray) and post (green) eye velocities in rightward single-target motion trials binned by time of measurement since motion onset.The center of the analysis window is used as the time of measurement.B, Same as A for Animal 1, negative step trials, motion condition is leftward (n = 98).C, Same as A for Animal 2, negative step trials, motion condition is downward (n = 234).D, Mean velocities in the direction of target motion for different target directions for pre-(gray) and post-saccadic (green) pursuit for Animal 1, positive step.Error bars are 1 SEM.E, Same as D for Animal 1, negative step.F, Same as D for Animal 2, negative step.

Figure 3 .
Figure 3. Two-target task examples.A, The scheme for two-target step-ramp pursuit task.Animals begin by fixating at a target at center of the screen.The targets then make either a positive or negative step (only negative step is shown) and start to move.After a variable interval, unselected target disappears.B-E, Example pursuit traces from animals.Motion of the target moving horizontally is shown in red and motion of target moving vertically is shown in blue.Eye movement is shown in black.Shaded regions in velocity plots indicate windows used to compute pre-(gray) and post-saccadic (green) eye velocities.Target steps and choice is indicated in schematic for each example.Weights for each example are provided.B and C are from Animal 1, C and D are from Animal 2.

Figure 4 .
Figure 4. Post-saccadic enhancement of selected target pursuit.A, Pre-(gray) and post-saccadic (green) eye positions for trials within each condition.B, Compares pre-(gray) and post-saccadic (green) pursuit weights to saccade weights.Thick lines indicate the fitted linear regression lines.C, ROC curves for pre-(gray) and post-saccadic pursuit weights.First column in all rows is data for Animal 1, positive steps, targets are moving along 0 and 90°(n = 401).Second column in all rows is data for Animal 1, negative step, targets are moving along 0 and 90°(n = 97).Third column in all rows is data for Animal 2, negative step, targets are moving along 180 and 90°(n = 301).Fourth column in all rows is data for Animal 2, negative step, targets are moving along 0 and 270°(n = 198).

Figure 5 .
Figure 5. Target selection evolution in time and relation to saccade.A, Weight distributions for pre-saccadic pursuit (gray), post-saccadic pursuit (green) and saccades (yellow).Animal, step, and direction conditions are indicated above the histograms.B, Choice probability in different intervals surrounding the saccade.Error bars are 95% confidence intervals on choice probability (n: first column = 401, second column = 97, third column = 198, fourth column = 301).

Figure 7 .
Figure 7. Scheme of target selection in the marmoset saccade and pursuit system.

Table 1 .
Effect of saccade on pursuit velocity

Table 2 .
Summary of relation between pursuit and saccade weights Primacy of spatial information in guiding target selection for pursuit and saccades.Available at: http://journalofvision.org/2/9/5/627 Beutter BR, Stone LS (1998) Human motion perception and smooth eye movements show similar directional biases for elongated apertures.Vision Res 38:1273-1286.