Reaction times of vertical prosaccades and antisaccades in gap and overlap tasks

Horizontal saccadic reaction times (SRTs) have been extensively studied over the past 3 decades, concentrating on such topics as the gap effect, express saccades, training effects, and the role of fixation and attention. This study investigates some of these topics with regard to vertical saccades. The reaction times of vertical saccades of 13 subjects were measured using the gap and the overlap paradigms in the prosaccade task (saccade to the stimulus) and the antisaccade task (saccade in the direction opposite to the stimulus). In the gap paradigm, the initial fixation point (FP) was extinguished 200 ms before stimulus onset, while, in the overlap paradigm, the FP remained on during stimulus presentation. With the prosaccade overlap task, it was found that most subjects (10/13) — whether they were previously trained making horizontal saccades or naive — had significantly faster upward saccades compared with their downward saccades. One subject was faster in the downward direction and two were symmetrical. The introduction of the gap reduced the reaction times of the prosaccades, and express saccades were obtained in some naive and most trained subjects. This gap effect was larger for saccades made to the downward target. The strength of the updown asymmetry was more pronounced in the overlap as compared to the gap paradigm. With the antisaccade task, up-down asymmetries were much reduced. Express antisaccades were absent even with the gap paradigm, but reaction times were reduced as compared to the antisaccade overlap paradigm. There was a slight tendency for a larger gap effect of downward saccades. All subjects produced a certain number of erratic prosaccades in the antitaks, more with the gap than with the overlap paradigm. There was a significantly larger gap effect for the erratic prosaccades made to the downward, as compared to the upward, target, due to increased downward SRTs in the overlap paradigm. Three subjects trained in both the horizontal and the vertical direction showed faster SRTs and more express saccades in the horizontal directions as compared to the vertical. It is concluded that different parts of the visual field are differently organized with both directional and nondirectional components in saccade preparation.     

Oculo-manual coordination control: respective role of visual and non-visual information in ocular tracking of self-moved targets

We evaluated the role of visual and non-visual information in the control of smooth pursuit movements during tracking of a self-moved target. Previous works have shown that self-moved target tracking is characterised by shorter smooth pursuit latency and higher maximal velocity than eye-alone tracking. In fact, when a subject tracks a visual target controlled by his own arm, eye movement and arm movement are closely synchronised. In the present study, we showed that, in a condition where the direction of motion of a self-moved visual target was opposite to that of the arm (same amplitude, same velocity, but opposite direction of movement), the resulting smooth pursuit eye movements occurred with low latency, and continued for about 140 ms in the direction of the arm movement rather than in the direction of the actual visual target movement. After 140 ms, the eye movement direction reversed through a combination of smooth pursuit and saccades. Subsequently, while arm and visual target still moved in opposite directions, smooth pursuit occurred in pace with the visual target motion. Subjects were also submitted to a series of 60 tracking trials, for which the arm-to-target motion relationship was systematically reversed. Under these conditions subjects were able to initiate early smooth pursuit in the actual direction of the visual target. Overall, these results confirm that non-visual information produced by the arm motor system can trigger and control smooth pursuit. They also demonstrate the plasticity of the neuronal network handling eye-arm coordination control.     

Express saccades in cat: effects of task and target modality

Saccadic eye movements to visual, auditory, and bimodal targets were measured in four adult cats. Bimodal targets were visual and auditory stimuli presented simultaneously at the same location. Three behavioral tasks were used: a fixation task and two saccadic tracking tasks (gap and overlap task). In the fixation task, a sensory stimulus was presented at a randomly selected location, and the saccade to fixate that stimulus was measured. In the gap and overlap tasks, a second target (hereafter called the saccade target) was presented after the cat had fixated the first target. In the gap task, the fixation target was switched off before the saccade target was turned on; in the overlap task, the saccade target was presented before the fixation target was switched off. All tasks required the cats to redirect their gaze toward the target (within a specified degree of accuracy) within 500 ms of target onset, and in all tasks target positions were varied randomly over five possible locations along the horizontal meridian within the cat's oculomotor range. In the gap task, a significantly greater proportion of saccadic reaction times (SRTs) were less than 125 ms, and mean SRTs were significantly shorter than in the fixation task. With visual targets, saccade latencies were significantly shorter in the gap task than in the overlap task, while, with bimodal targets, saccade latencies were similar in the gap and overlap tasks. On the fixation task, SRTs to auditory targets were longer than those to either visual or bimodal targets, but on the gap task, SRTs to auditory targets were shorter than those to visual or bimodal targets. Thus, SRTs reflected an interaction between target modality and task. Because target locations were unpredictable, these results demonstrate that cats, as well as primates, can produce very short latency goal-directed saccades.     

An eye-to-hand magnet effect reveals distinct spatial interference in motor planning and execution

An important question in oculomanual control is whether motor planning and execution modulate interference between motion of the eyes and hands. Here we investigated oculomanual interference using a novel paradigm that required saccadic eye movements and unimanual finger tapping. We examined finger trajectories for spatial interference caused by concurrent saccades. The first experiment used synchronous cues so that saccades and taps shared a common timekeeping goal. We found that finger trajectories showed bilateral interference where either finger was attracted in the direction of the accompanying saccade. The second experiment avoided interference due to shared planning resources by examining interference caused by reactive saccades. Here, we observed a lesser degree of execution-dependent coupling where the finger trajectory deviated only when reactive saccades were directed toward the hemifield of the responding hand. Our results show that distinct forms of eye-to-hand coupling emerge according to the demands of the task.     

Centripetal bias on preparation for smooth pursuit eye movements based on the anticipation

It has been reported that a brief perturbation of a stationary target during fixation induces larger eye movement when monkeys anticipate future smooth pursuit than when they do not. Here, we recorded eye movements in human subjects after briefly perturbing a target and the eccentricity of its initial position was changed under three conditions: (1) subjects anticipated saccades for a target that appeared before; (2) they anticipated smooth pursuit for a target that appeared before; and (3) they anticipated smooth pursuit but did not know beforehand where the target started from. We found that in condition 2 substantial eye movements were induced by the perturbation started moving toward the center. However, weak responses were observed in conditions 1 and 3. These results indicate that ocular responses to brief perturbations of the target at eccentric positions are increased with centripetal bias when human subjects prepare for future smooth pursuit.     

Latency of saccades during smooth-pursuit eye movement in man Directional asymmetries

To examine the effects of smooth-pursuit eye movements on the initiation of saccades, their latency was measured when subjects initially fixated or pursued a target. In half of the block of trials, the fixation or pursuit target was extinguished 200 ms before the saccade target was illuminated (gap trials). Reduction of the mean saccade latency in the gap trials (the “gap effect”) was evident even when the subjects were pursuing a moving target, consistent with previous observations. The effect of pursuit direction on saccade latency was also examined. Saccades in the same direction as the preceding pursuit (forward saccades) had shorter latencies than those in the opposite direction (backward saccades). This asymmetry was observed in both the gap and nongap trials. Although the forward-backward asymmetry was much smaller than the “gap effect”, it was statistically significant in six of eight cases. These results suggest that the preparation of saccades is affected by smooth-pursuit eye movements. Received: 2 June 1997 / Accepted: 6 November 1997     

Coordination of smooth pursuit and saccade target selection in monkeys

The coordination of saccadic and smooth pursuit eye movements in macaque monkeys was investigated using a target selection paradigm with two moving targets crossing at a center fixation point. A task in which monkeys selected a target based on its color was used to test the hypothesis that common neural signals underlie target selection for pursuit and saccades, as well as testing whether target selection signals are available to the saccade and pursuit systems simultaneously or sequentially. Several combinations of target color, speed, and direction were used. In all cases, smooth pursuit was highly selective for the rewarded target before any saccade occurred. On >80% of the trials, the saccade was directed toward the same target as both pre- and postsaccadic pursuit. The results favor a model in which a shared target selection signal is simultaneously available to both the saccade and pursuit systems, rather than a sequential model.     

Effects of dual task demands on the accuracy of smooth pursuit eye movements

The effect of attention allocation on smooth pursuit eye movements (SPEM) was investigated. Eye movements were electrooculographically recorded in 27 healthy subjects who tracked a visual target that moved horizontally with constant or unpredictably varying velocity. In some trials, subjects performed additional auditory discrimination tasks varying in difficulty. Pursuit error decreased when attention was divided between both tasks. The pattern of results is incompatible with the assumption made in previous research that attention enhancement improves SPEM accuracy. Rather, ocular smooth pursuit appears to be executed in the automatic mode, although intentional and selective processes must contribute. Moreover, controlled attention directed to the tracking task interfered with smooth pursuit. A reinterpretation of earlier studies in which visual monitoring tasks were used to improve eye tracking is needed.     

Adaptation of catch-up saccades during the initiation of smooth pursuit eye movements

Reduction of retinal speed and alignment of the line of sight are believed to be the respective primary functions of smooth pursuit and saccadic eye movements. As the eye muscles strength can change in the short-term, continuous adjustments of motor signals are required to achieve constant accuracy. While adaptation of saccade amplitude to systematic position errors has been extensively studied, we know less about the adaptive response to position errors during smooth pursuit initiation, when target motion has to be taken into account to program saccades, and when position errors at the saccade endpoint could also be corrected by increasing pursuit velocity. To study short-term adaptation (250 adaptation trials) of tracking eye movements, we introduced a position error during the first catch-up saccade made during the initiation of smooth pursuit—in a ramp-step-ramp paradigm. The target position was either shifted in the direction of the horizontally moving target (forward step), against it (backward step) or orthogonally to it (vertical step). Results indicate adaptation of catch-up saccade amplitude to back and forward steps. With vertical steps, saccades became oblique, by an inflexion of the early or late saccade trajectory. With a similar time course, post-saccadic pursuit velocity was increased in the step direction, adding further evidence that under some conditions pursuit and saccades can act synergistically to reduce position errors.     

Ocular pursuit and the estimation of time-to-contact with accelerating objects in prediction motion are controlled independently based on first-order estimates

The present study examined for the first time both the ocular and manual responses in a prediction motion (PM) task requiring participants to estimate time-to-contact (TTC) of an accelerating object. Results showed that while the ocular response initially matched well the object motion, smooth pursuit decayed towards zero following object occlusion, during which participants exhibited a saccadic response that placed the eyes in the region of the point of contact. The primary saccade was completed in advance of the object reaching the point of contact, and was best predicted by a first-order estimate of TTC (TTC1). Participants then made their manual response, which was also best predicted by TTC1. Therefore, object acceleration was not taken into account in either the ocular or manual response, with the latter occurring before the object reached the point of contact when it decelerated and after when it accelerated. Further analyses of the ocular and manual responses indicated no functional relationship and independent control. We suggest that the demand to make temporal estimates with a stationary location in PM tasks is critical in explaining the discrepancy with oculomotor research.     

Quantitative analysis of catch-up saccades during sustained pursuit

During visual tracking of a moving stimulus, primates orient their visual axis by combining two very different types of eye movements, smooth pursuit and saccades. The purpose of this paper was to investigate quantitatively the catch-up saccades occurring during sustained pursuit. We used a ramp-step-ramp paradigm to evoke catch-up saccades during sustained pursuit. In general, catch-up saccades followed the unexpected steps in position and velocity of the target. We observed catch-up saccades in the same direction as the smooth eye movement (forward saccades) as well as in the opposite direction (reverse saccades). We made a comparison of the main sequences of forward saccades, reverse saccades, and control saccades made to stationary targets. They were all three significantly different from each other and were fully compatible with the hypothesis that the smooth pursuit component is added to the saccadic component during catch-up saccades. A multiple linear regression analysis was performed on the saccadic component to find the parameters determining the amplitude of catch-up saccades. We found that both position error and retinal slip are taken into account in catch-up saccade programming to predict the future trajectory of the moving target. We also demonstrated that the saccadic system needs a minimum period of approximately 90 ms for taking into account changes in target trajectory. Finally, we reported a saturation (above 15 degrees /s) in the contribution of retinal slip to the amplitude of catch-up saccades.     

The initiation of smooth pursuit eye movements and saccades in normal subjects and in “express-saccade makers”

A vast knowledge exists about saccadic reaction times (RT) and their bi- or multimodal distributions with very fast (express) and regular RT. Recently, there has been some evidence that the smooth pursuit system may show a similar RT behavior. Since moving targets usually evoke a combined pursuit/saccade response, we asked which processes influence the initiation of pursuit and saccadic eye movements. Furthermore, we investigated whether and how the pursuit and saccadic system interact during the initiation of eye movements to moving targets. We measured the RT of the initial smooth pursuit (iSP) response and of the first corrective saccade and compared the RT behavior of both. Furthermore we compared the behavior of the corrective saccades to moving targets to that of saccades to stationary targets, known from the literature. The stimulus consisted of a target that moved suddenly at constant velocity (ramp). In addition, prior to the movement, a temporal gap, a position step or a combination of both could occur (gap-ramp, step-ramp, gap-step-ramp, respectively). Differently from most previous studies, we chose step and ramp with the same direction to provoke competition between the pursuit and saccade system. For the first time we investigated pursuit initiation in "express-saccade makers" (ES makers), a subject group known to produce an abnormally high percentage of short-latency saccades in saccade tasks. We compared their results with subject groups who were either naive or trained with respect to saccade tasks. The iSP started at approximately 100 ms, which corresponds to express saccade latencies. These short iSP-RT occurred reflex-like and almost independent of the experimental task. A bimodal frequency distribution of RT with a second peak of longer iSP-RT occurred exclusively in the ramp paradigm. The RT of the first corrective saccades in a pursuit task were comparable with that in a saccade task and depended on the stimulus. The ability of ES makers to produce a high number of express saccades was transferred to corrective saccades in the pursuit task, but not to pursuit initiation. In summary, short-latency pursuit responses differ from express saccades with respect to their independence of experiment and subject group. Therefore, a simple analogy to express saccades cannot be drawn, although some mechanisms seem to act similarly on both the pursuit and the saccade system (such as disengagement of attention with the gap effect). Furthermore, we found evidence that the initial pursuit response and the first corrective saccade are processed independently of each other. The first corrective saccades to moving targets behave like saccades to stationary targets. Normal pursuit but abnormal saccade RT of ES makers can be explained by recent theories of superior colliculus (SC) function in terms of retinal error handling.     

Role of retinal slip in the prediction of target motion during smooth and saccadic pursuit

Visual tracking of moving targets requires the combination of smooth pursuit eye movements with catch-up saccades. In primates, catch-up saccades usually take place only during pursuit initiation because pursuit gain is close to unity. This contrasts with the lower and more variable gain of smooth pursuit in cats, where smooth eye movements are intermingled with catch-up saccades during steady-state pursuit. In this paper, we studied in detail the role of retinal slip in the prediction of target motion during smooth and saccadic pursuit in the cat. We found that the typical pattern of pursuit in the cat was a combination of smooth eye movements with saccades. During smooth pursuit initiation, there was a correlation between peak eye acceleration and target velocity. During pursuit maintenance, eye velocity oscillated at approximately 3 Hz around a steady-state value. The average gain of smooth pursuit was approximately 0.5. Trained cats were able to continue pursuing in the absence of a visible target, suggesting a role of the prediction of future target motion in this species. The analysis of catch-up saccades showed that the smooth-pursuit motor command is added to the saccadic command during catch-up saccades and that both position error and retinal slip are taken into account in their programming. The influence of retinal slip on catch-up saccades showed that prediction about future target motion is used in the programming of catch-up saccades. Altogether, these results suggest that pursuit systems in primates and cats are qualitatively similar, with a lower average gain in the cat and that prediction affects both saccades and smooth eye movements during pursuit.     

The influence of stimulus direction and eccentricity on pro- and anti-saccades in humans

We examined the sensory and motor influences of stimulus eccentricity and direction on saccadic reaction times (SRTs), direction-of-movement errors, and saccade amplitude for stimulus-driven (prosaccade) and volitional (antisaccade) oculomotor responses in humans. Stimuli were presented at five eccentricities, ranging from 0.5° to 8°, and in eight radial directions around a central fixation point. At 0.5° eccentricity, participants showed delayed SRT and increased direction-of-movement errors consistent with misidentification of the target and fixation points. For the remaining eccentricities, horizontal saccades had shorter mean SRT than vertical saccades. Stimuli in the upper visual field trigger overt shifts in gaze more easily and faster than in the lower visual field: prosaccades to the upper hemifield had shorter SRT than to the lower hemifield, and more anti-saccade direction-of-movement errors were made into the upper hemifield. With the exception of the 0.5° stimuli, SRT was independent of eccentricity. Saccade amplitude was dependent on target eccentricity for prosaccades, but not for antisaccades within the range we tested. Performance matched behavioral measures described previously for monkeys performing the same tasks, confirming that the monkey is a good model for the human oculomotor function. We conclude that an upper hemifield bias lead to a decrease in SRT and an increase in direction errors.     

Differential effects of blinks on horizontal saccade and smooth pursuit initiation in humans

Blinks executed during eye movements affect kinetic eye movement parameters, e.g., peak velocity of saccades is decreased, their duration is increased, but their amplitude is not altered. This effect is mainly explained by the decreased activity of premotor neurons in the brainstem: omni-pause neurons (OPN) in the nucleus raphe interpositus. Previous studies examined the immediate effect of blinks directly on eye movements but not their effect when they are elicited several hundred milliseconds before the eye movements. In order to address this question we tested blinks elicited before the target onset of saccades and pursuit and compared the results to the gap effect: if a fixation light is extinguished for several hundred milliseconds, the reaction time (latency) for subsequent saccades or smooth pursuit eye movements is decreased. Monocular eye and lid movements were recorded in nine healthy subjects using the scleral search-coil system. Laser stimuli were front-projected onto a tangent screen in front of the subjects. Horizontal step-ramp smooth pursuit of 20 deg/s was elicited in one session, or 5 deg horizontal visually guided saccades in another experimental session. In one-third of the trials (smooth pursuit or saccades) the fixation light was extinguished for 200 ms before stimulus onset (gap condition), and in another third of the trials reflexive blinks were elicited by a short airpuff before the stimulus onset (blink condition). The last third of the trials served as controls (control condition). Stimulus direction and the three conditions were randomized for saccades and smooth pursuit separately. The latency of the step-ramp smooth pursuit in the blink condition was found to be decreased by 10 ms, which was less than in the gap condition (38 ms). However, the initial acceleration and steady-state velocity of smooth pursuit did not differ in the three conditions. In contrast, the latency of the saccades in the gap condition was decreased by 39 ms, but not in the blink condition. Saccade amplitude, peak velocity, and duration were not different in the three conditions. There was also no difference in blink amplitude and duration of pupil occlusion in the blink condition, neither in saccades nor in smooth pursuit. The latency reduction of smooth pursuit, but not of saccades, may neither be explained by the brief pupil occlusion nor by visual suppression, warning signals, or the startle response. Whether the effects are caused by the influence of blinks on OPNs or other premotor structures remains to be tested.     

Facilitation of smooth pursuit initiation by electrical stimulation in the supplementary eye fields.

The role of the supplementary eye fields (SEF) during smooth pursuit was investigated with electrical microstimulation. We found that stimulation in the SEF increased the acceleration and velocity of the eyes in the direction of target motion during smooth pursuit initiation but not during sustained pursuit. The increase in eye velocity during initiation will be referred to as pursuit facilitation and was observed at sites where saccades could not be evoked with the same stimulation parameters. On average, electrical stimulation increased eye velocity by approximately 20%. At most sites, the threshold for a significant facilitation was 50 microA with a stimulation frequency of 300 Hz. Facilitation of pursuit initiation depended on the timing of stimulation trains. The effect was most pronounced if the stimulation was delivered before smooth pursuit initiation. On average, eye velocity in stimulation trials increased linearly as a function of eye velocity in control trials, and this function had a slope greater than one, suggesting a multiplicative influence of the stimulation. Stimulation during a fixation task did not evoke smooth eye movements. The latency of catch-up saccades was increased during facilitation, but their accuracy was not affected. Saccades toward stationary targets were not affected by the stimulation. The results are further evidence that the SEF plays a role in smooth pursuit in addition to its known role in saccade planning and suggest that this role may be to control the gain of smooth pursuit during initiation. The covariance between pursuit facilitation and the timing of the catch-up saccade as a result of stimulation suggests that these different eye movements systems are coordinated to achieve a common goal.     

Effects of pre-cues on voluntary and reflexive saccade generation I. Anti-cues for pro-saccades

Experiments on visual attention have employed both physical cues and verbal instructions to enable subjects to allocate attention at a location that becomes relevant within a perceptual or motor task some time later (cue lead time, CLT). In this study we have used valid visual peripheral cues (CLT between 100 and 700 ms) to indicate the direction and location of the next saccade. A cue is considered valid or invalid if its meaning with respect to the next saccade is correct or incorrect. A cue is called an anti- or pro-cue if the side of its presentation is opposite to or the same as the direction of the saccade required on a given trial. Correspondingly, a saccade is called an anti- or pro-saccade if it is directed to the side opposite to or the same as the stimulus presentation. A condition in which the cue and the stimulus are presented on opposite sides provides a simple way of dissociating voluntary attention allocation from automatic orienting. This paper considers the anti-cue pro-saccade task: the subjects were instructed to use the cue to direct attention to the opposite side, i.e. the location, where on valid trials the saccade target would occur. In the companion paper we have used the same physical condition, but we have reversed the instructions as to saccade direction and we have reversed the meaning of the cue, i.e. we designed a pro-cue anti-saccade task. In this first paper, the saccadic reaction times (SRTs) of pro-saccades of five adult subjects were measured in the gap paradigm (fixation point offset precedes target onset by 200 ms). With a CLT of 100 ms, valid anti-cues reduced the number of express saccades (i.e. saccades with SRTs in the range 80–120 ms) significantly compared with the control values (no cues). Valid anti-cues with increasingly long CLTs (100–700 ms) resulted in an increasing incidence of anticipatory saccades and saccades with longer SRTs (more than 120 ms), while the frequency of express saccades remained below the control value. When cue and saccade target were dissociated in location or in both location and direction, the effects of the cueing revealed a much lower spatial selectivity as compared to the effects that have been described for voluntary attention allocation by means of central cues. The results suggest that voluntary allocation of attention and cue-induced automatic orienting not only have different time courses but also have opposite effects on the generation of express saccades, and different spatial selectivities. A possible neuronal basis of these results is discussed considering related findings from electrophysiological studies in monkeys. Received: 27 March 1997 / Accepted: 17 December 1997     

Oculomotor tracking in two dimensions

Results from studies of oculomotor tracking in one dimension have indicated that saccades are driven primarily by errors in position, whereas smooth pursuit movements are driven primarily by errors in velocity. To test whether this result generalizes to two-dimensional tracking, we asked subjects to track a target that moved initially in a straight line then changed direction. We found that the general premise does indeed hold true; however, the study of oculomotor tracking in two dimensions provides additional insight. The first saccade was directed slightly in advance of target location at saccade onset. Thus its direction was related primarily to angular positional error. The direction of the smooth pursuit movement after the saccade was related linearly to the direction of target motion with an average slope of 0.8. Furthermore the magnitude and direction of smooth pursuit velocity did not change abruptly; consequently the direction of smooth pursuit appeared to rotate smoothly over time.     

Stimulus intensity modifies saccadic reaction time and visual response latency in the superior colliculus

Performance in a reaction time task can be strongly influenced by the physical properties of the stimuli used (e.g., position and intensity). The reduction in reaction time observed with higher-intensity visual stimuli has been suggested to arise from reduced processing time along the visual pathway. If this hypothesis is correct, activity should be registered in neurons sooner for higher-intensity stimuli. We evaluated this hypothesis by measuring the onset of neural activity in the intermediate layers of the superior colliculus while monkeys generated saccades to high or low-intensity visual stimuli. When stimulus intensity was high, the response onset latency was significantly reduced compared to low-intensity stimuli. As a result, the minimum time for visually triggered saccades was reduced, accounting for the shorter saccadic reaction times (SRTs) observed following high-intensity stimuli. Our results establish a link between changes in neural activity related to stimulus intensity and changes to SRTs, which supports the hypothesis that shorter SRTs with higher-intensity stimuli are due to reduced processing time.     

Task-dependent effects of social attention on saccadic reaction times

Previous research has shown that saccadic reaction times (SRTs) are shorter when a stimulus is flashed on the same side as the observed gaze direction of another individual. The gaze imitation hypothesis contends that observed gaze evokes the preparation of a saccade toward the same direction. Previous studies of this phenomenon have employed pro-saccade tasks in which the instructed saccade is directed toward the stimulus. In agreement with previous findings, we found that SRTs on pro-saccade trials were shorter when the stimulus appeared in the same direction as observed gaze. Here we also included anti-saccade trials in which subjects were required to look-away from a stimulus and toward its mirror position in the opposite visual field. The gaze imitation hypothesis predicts that subjects will have shorter SRTs on anti-saccade trials in which the stimulus appears opposite the observed gaze direction because they will have prepared already a saccade in that direction. However, contrary to the prediction of the gaze imitation hypothesis, we found that subjects had shorter SRTs on anti-saccade trials when the stimulus appeared in the same direction as observed gaze. Moreover, subjects also made more pro-saccade errors on anti-saccade trials in which the stimulus was presented opposite the observed gaze direction. The results of our study indicate that subjects prepared a saccade in the same direction as observed gaze on pro-saccade trials but opposite the observed gaze direction on anti-saccade trials. These findings suggest that the effect of social gaze cues on SRTs is task dependent.